{"schema":"coreai-decider-vision-swift-gate/1","gate":"Swift read-out (apps/DeciderVision, decider-vision CLI) vs the author's fp32 oracle, the Python host reference and the Python read-out of the same bundle","redacted":{"why":"the GPU-lock snapshots listed every user process with its full command line (home paths, session ids, another lane's files); a public transcript keeps only each process's executable name, pid, CPU, RSS and age","full_transcript":"$ZOO_WORK_ROOT/_decider2bv/transcripts_full/gate-decider-2b-vision-swift.full.json","full_sha256":"e22585d56f07f203ef2c5c416d7b4769db0e1c1e5614265890f3a44410d80689","full_bytes":549871,"commands_reduced_to_executable_name":247,"redacted_on":"2026-09-29"},"oracle":{"path":"/Users/majimadaisuke/code/coreai/_decider2bv/oracle/fixture_oracle.json","sha256":"c2741795879ee1036fb1b4e3b9c016174fc3a4de90b448d9198718a39b01573e","rows_json_sha256":"b3dd44b1e2115cdc8b80ede88ed8a2c1addf341a79eef55ffaa784cb6a4323ce","runs":76,"slots":108},"package":{"path":"/Users/majimadaisuke/code/coreai/coreai-models-community-decider2bv-wt/apps/DeciderVision","sources_sha256":{"Package.swift":"ff4cb5d1b65260e657962666b53a862750b0b7566d656603be043283c39cbcc2","Sources/DeciderVision/DecisionDecoder.swift":"b5cdf022598bcc42516789a005a2d99e84525a0fc42db7debdef5fcaa9c157f1","Sources/DeciderVision/ImagePreprocess.swift":"e09d6c1ef707b428642ae4699e153edb71ce8f7565d85c6628c1be856889f07d","Sources/DeciderVision/PromptBuilder.swift":"c703f7693b981454ccb83e346fc7d093b0834aa945596ec68deb1be890aea7a9","Sources/DeciderVision/Support.swift":"483697ecc4adcc9433ee92a329ace5c782392b869720d086f083274b2be86fb3","Sources/DeciderVision/VisionDecider.swift":"ec2539633f947a2b93101ef7b3c0189b2cf8a041515d1c14c8aeffa741fe8a1a","Sources/DeciderVision/VisionTower.swift":"f5a691134c70ff2337b58662eef4d4fe2509b72561f1560db73ee53fc8e6dc90","Sources/decider-vision/main.swift":"ff04620a16a2e0a95cc0ec74511710a38dc92c3808c4d8a8dfcc45f436926699"},"resolved":{"eventsource":"1.5.1","swift-asn1":"1.7.3","swift-collections":"1.7.1","swift-crypto":"4.5.2","swift-huggingface":"0.11.0","swift-jinja":"2.5.1","swift-transformers":"1.3.4","yyjson":"0.12.0"}},"s1":{"runs_expected":76,"ids_equal_oracle":76,"slots_equal_oracle":76,"image_runs_rope_and_block_equal":70,"image_runs":70,"extra_runs":[],"failures":[],"result":"PASS","definition":"Swift ids with V+k mapped to <|image_pad|> = fixture_oracle.json ids token for token; slots = slot_idx; image runs: rope_shift_start = 1 + G^2, rope_shift_amount = the oracle's rope_shift, ids[1 : 1 + G^2] = V + 0 .. V + G^2 - 1; text runs: start 1 << 30, amount 0"},"s2":{"tiles":58,"tiles_expected":58,"max_abs_level":0,"worst":{"image":"mps_red","grid":"g256","max_abs_level_vs_resize_bicubic":0,"where_yxc":null},"tiles_bit_equal":58,"decoded_rgb_equal_meta":58,"patches_sha256_equal_host":58,"vs_pillow_values_only":{"max_abs_level":2,"worst":{"image":"two_shapes_640x360","grid":"g448","max_abs_level_vs_pillow":2},"tiles_bit_equal":56},"bar":"max |level| <= 1 vs host.resize_bicubic (expected 0); decoded RGB = meta rgb_sha256","result":"PASS","tiles_detail":[{"image":"mps_red","grid":"g256","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"mps_red","grid":"g448","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"mps_green","grid":"g256","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"mps_green","grid":"g448","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"mps_blue","grid":"g256","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"mps_blue","grid":"g448","size_in":[224,224],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"circle_red_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"circle_red_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"square_blue_320x240","grid":"g256","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"square_blue_320x240","grid":"g448","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"triangle_green_240x320","grid":"g256","size_in":[240,320],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"triangle_green_240x320","grid":"g448","size_in":[240,320],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"star_yellow_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"star_yellow_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"dots_three_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"dots_three_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"dots_five_320x240","grid":"g256","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"dots_five_320x240","grid":"g448","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"ball_left_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"ball_left_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"stripes_vertical_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"stripes_vertical_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"checker_320x240","grid":"g256","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"checker_320x240","grid":"g448","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"grid_cell_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"grid_cell_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"arrow_up_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"arrow_up_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"arrow_left_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"arrow_left_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"bars_320x240","grid":"g256","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"bars_320x240","grid":"g448","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"lamp_top_240x320","grid":"g256","size_in":[240,320],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"lamp_top_240x320","grid":"g448","size_in":[240,320],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"bands_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"bands_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"two_shapes_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"two_shapes_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":2,"n_differing_vs_pillow":558},{"image":"seg_7_256","grid":"g256","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"seg_7_256","grid":"g448","size_in":[256,256],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"seg_3_320x240","grid":"g256","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"seg_3_320x240","grid":"g448","size_in":[320,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"seg_42_640x360","grid":"g256","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"seg_42_640x360","grid":"g448","size_in":[640,360],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":1,"n_differing_vs_pillow":166},{"image":"pong_up_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"pong_up_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"pong_down_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"pong_down_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"pong_stay_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"pong_stay_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"breakout_left_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"breakout_left_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"breakout_launch_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"breakout_launch_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"minigrid_forward_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"minigrid_forward_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"cliff_up_256x240","grid":"g256","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0},{"image":"cliff_up_256x240","grid":"g448","size_in":[256,240],"decode_path":"direct RGBX","decoded_rgb_sha256_equals_meta":true,"max_abs_level_vs_resize_bicubic":0,"where_yxc":null,"n_channel_values_differing":0,"patches_sha256_equals_host":true,"max_abs_level_vs_pillow":0,"n_differing_vs_pillow":0}]},"s3":{"asset":"aot","bundle":"decider_2b_vision_decode_int8mix_pf16","fixture_json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/aot/fixture_aot.json","label":"S1-S3 AOT int8mix_pf16 (correctness run, final binary, GPU shared with round 7)","assets":{"asset":"aot","bundle":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles/decider_2b_vision_decode_int8mix_pf16","decoder":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles_aotc/decider_2b_vision_decode_int8mix_pf16.h16c.aimodelc","decoder_options":"SpecializationOptions.default","tower_options":"SpecializationOptions.default"},"environment":{"argv":["swift/.build/release/decider-vision","fixture","--bundle","exports/bundles/decider_2b_vision_decode_int8mix_pf16","--tower-g256","exports/decider_2b_vision_g256_vision_fp16w32_aotc/decider_2b_vision_g256_vision_fp16w32.h16c.aimodelc","--tower-g448","exports/decider_2b_vision_g448_vision_fp16w32_aotc/decider_2b_vision_g448_vision_fp16w32.h16c.aimodelc","--rows","fixtures/rows.json","--images","fixtures/images","--meta","fixtures/meta.json","--out","swift/aot/fixture_aot.json","--label","S1-S3 AOT int8mix_pf16 (correctness run, final binary, GPU shared with round 7)"],"build_configuration":"Release","chip":"Apple M4 Max","device_architecture":"h16c","model":"Mac16,9","os":"Version 27.0 (Build 26A428)","os_build":"26A428","pid":81133},"load_first":{"decoder_s":{"main":3.942049666,"model":0.0014457080000000002,"prefill":16.693806958},"tokenizer_s":0.792190166,"tower_s":{"g256":0.901461292,"g448":0.9710682920000001},"wall_s":23.306870417},"function_names":["main","prefill"],"reset_check":{"bit_equal":true,"run":"r01/g256","slot_logits_max_abs_diff":0,"wall_from_file_s":0.301698792},"bar":{"max_abs_dp":0.02,"mean_of_run_mean_abs_dp":0.002,"source":"readout_gate_vision.BAR (round 4, unchanged)","slots":"all","argmax":"all","full_vocab_top1":"the oracle's argmax letter on every slot","reset":"bit-equal re-run"},"summary":{"runs":76,"runs_expected":76,"slots_read":108,"slots_expected":108,"argmax_equal":108,"full_vocab_top1_is_oracle_letter":108,"max_abs_dp":0.009645075075595422,"worst":{"id":"r26","arm":"g448","t":276,"max_abs_dp":0.009645075075595422,"probs":[0.8428485100197483,0.02975619181099293,0.0854323417771534,0.04196295639210546],"probs_oracle":[0.8332034349441528,0.030570058152079582,0.08590248227119446,0.050323981791734695]},"mean_of_run_mean_abs_dp":0.00027963630186201313,"finite_all":true,"reset_bit_equal":true,"swift_json_probs_equal_recomputed":108,"slots_read_from_prefill":9},"checks":{"all_runs":true,"slots":true,"argmax":true,"full_vocab_top1":true,"max_abs_dp":true,"mean_of_run_mean_abs_dp":true,"finite":true,"reset_bit_equal":true},"result":"PASS","per_arm":{"g256":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.009505770333302022,"mean_of_run_mean_abs_dp":0.00028210424415138713},"g448":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.009645075075595422,"mean_of_run_mean_abs_dp":0.00029473326142662935},"text":{"runs":6,"slots":10,"argmax_equal":10,"full_vocab_top1_is_oracle_letter":10,"max_abs_dp":0.0019238409545840884,"mean_of_run_mean_abs_dp":0.00017717437438040406}},"vs_python_same_bundle":{"definition":"per slot, the same bundle read by Python: image runs vs the round-6 e2e arm (decider_2b_vision_decode_int8mix_pf16+tower_fp16w32+chunk16: Pillow resize -> same tower .aimodelc -> same decoder .aimodelc, chunk order), text runs vs the round-6 b1 transcript (identical inputs). |dp| over the question's letters; bit-equality of the fp16 logits and of the tower's float32 output","all":{"slots":108,"max_abs_dp":2.8676610337496954e-05,"mean_abs_dp_of_slot_max":4.888431609297693e-07,"letter_logits_bit_equal":105,"full_logits_bit_equal":104,"full_logits_max_abs_diff":0.015625,"argmax_equal":108,"tower_embeds_bit_equal_runs":68,"tower_embeds_max_abs_diff":0.053189992904663086},"by_arm":{"g256":{"slots":49,"max_abs_dp":0.0,"mean_abs_dp_of_slot_max":0.0,"letter_logits_bit_equal":49,"full_logits_bit_equal":49,"full_logits_max_abs_diff":0.0,"argmax_equal":49,"tower_embeds_bit_equal_runs":35,"tower_embeds_max_abs_diff":0.0},"g448":{"slots":49,"max_abs_dp":2.8676610337496954e-05,"mean_abs_dp_of_slot_max":1.077450232253369e-06,"letter_logits_bit_equal":46,"full_logits_bit_equal":45,"full_logits_max_abs_diff":0.015625,"argmax_equal":49,"tower_embeds_bit_equal_runs":33,"tower_embeds_max_abs_diff":0.053189992904663086},"text":{"slots":10,"max_abs_dp":0.0,"mean_abs_dp_of_slot_max":0.0,"letter_logits_bit_equal":10,"full_logits_bit_equal":10,"full_logits_max_abs_diff":0.0,"argmax_equal":10,"tower_embeds_bit_equal_runs":0,"tower_embeds_max_abs_diff":0.0}}},"python_bundle_records":{"e2e":"b75795b701f2f3c5fc5af4905819d59c73f2872fb0a92889713fed5295aedaa4","b1":"b75795b701f2f3c5fc5af4905819d59c73f2872fb0a92889713fed5295aedaa4"},"runs":[{"id":"r01","arm":"g256","tokens":105,"max_abs_dp":2.5768715975482692e-06,"mean_abs_dp":1.7408191129301828e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r01","arm":"g448","tokens":237,"max_abs_dp":2.768114484186235e-06,"mean_abs_dp":1.8362613521086973e-06,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g256","tokens":105,"max_abs_dp":1.7517262937971623e-06,"mean_abs_dp":1.151955941633693e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g448","tokens":237,"max_abs_dp":3.5807912063257064e-06,"mean_abs_dp":2.3882127716399297e-06,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g256","tokens":105,"max_abs_dp":8.002206732693473e-06,"mean_abs_dp":5.321125688121196e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g448","tokens":237,"max_abs_dp":1.5796380499533136e-05,"mean_abs_dp":1.0553039244136881e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g256","tokens":153,"max_abs_dp":2.5983379817710525e-05,"mean_abs_dp":1.1894317126882495e-05,"calls":{"main":9,"prefill":9,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g448","tokens":285,"max_abs_dp":1.0543198452852415e-05,"mean_abs_dp":4.222475882120524e-06,"calls":{"main":13,"prefill":17,"total":30},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g256","tokens":105,"max_abs_dp":0.00014047266045336126,"mean_abs_dp":9.366885178863718e-05,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g448","tokens":237,"max_abs_dp":0.00017292879494146263,"mean_abs_dp":0.00011528733788843245,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g256","tokens":126,"max_abs_dp":4.116295669320813e-05,"mean_abs_dp":1.0282297243470957e-05,"calls":{"main":14,"prefill":7,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g448","tokens":258,"max_abs_dp":3.389932141206753e-05,"mean_abs_dp":8.480616558330334e-06,"calls":{"main":2,"prefill":16,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g256","tokens":184,"max_abs_dp":0.000555964786662852,"mean_abs_dp":0.00010669538210402103,"calls":{"main":24,"prefill":10,"total":34},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g448","tokens":316,"max_abs_dp":0.0007992280659719908,"mean_abs_dp":0.0001533721596194714,"calls":{"main":28,"prefill":18,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g256","tokens":120,"max_abs_dp":2.8015877093889863e-05,"mean_abs_dp":9.334585116129031e-06,"calls":{"main":8,"prefill":7,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g448","tokens":252,"max_abs_dp":2.9646596945598347e-05,"mean_abs_dp":9.889592567457425e-06,"calls":{"main":12,"prefill":15,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g256","tokens":128,"max_abs_dp":0.00019525454029145628,"mean_abs_dp":5.81961361324308e-05,"calls":{"main":0,"prefill":8,"total":8},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g448","tokens":260,"max_abs_dp":0.0012878015402648657,"mean_abs_dp":0.00032691769202498845,"calls":{"main":4,"prefill":16,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g256","tokens":112,"max_abs_dp":3.962290286470575e-06,"mean_abs_dp":2.6017031163013717e-06,"calls":{"main":0,"prefill":7,"total":7},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g448","tokens":244,"max_abs_dp":5.277150986970902e-06,"mean_abs_dp":3.5075189903071883e-06,"calls":{"main":4,"prefill":15,"total":19},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g256","tokens":103,"max_abs_dp":1.5927995652420535e-06,"mean_abs_dp":1.564277811323114e-06,"calls":{"main":7,"prefill":6,"total":13},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g448","tokens":235,"max_abs_dp":4.3411948593958735e-05,"mean_abs_dp":4.3387734207054295e-05,"calls":{"main":11,"prefill":14,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g256","tokens":117,"max_abs_dp":2.3122028144029194e-05,"mean_abs_dp":9.274080010546421e-06,"calls":{"main":5,"prefill":7,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g448","tokens":249,"max_abs_dp":1.3570069607782287e-05,"mean_abs_dp":5.424104646745695e-06,"calls":{"main":9,"prefill":15,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g256","tokens":140,"max_abs_dp":0.0010179375096165157,"mean_abs_dp":0.0002270132377130864,"calls":{"main":12,"prefill":8,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g448","tokens":272,"max_abs_dp":1.5331796330286807e-06,"mean_abs_dp":4.862470358999081e-07,"calls":{"main":0,"prefill":17,"total":17},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g256","tokens":111,"max_abs_dp":5.137983285918857e-05,"mean_abs_dp":2.5674156705415657e-05,"calls":{"main":15,"prefill":6,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g448","tokens":243,"max_abs_dp":2.0145859257691967e-05,"mean_abs_dp":1.0082197379877578e-05,"calls":{"main":3,"prefill":15,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g256","tokens":147,"max_abs_dp":6.602053493132143e-05,"mean_abs_dp":2.878435718025259e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g448","tokens":279,"max_abs_dp":0.0002839418392824866,"mean_abs_dp":0.000128595348353924,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g256","tokens":193,"max_abs_dp":7.170317945405383e-05,"mean_abs_dp":2.339074517043821e-05,"calls":{"main":17,"prefill":11,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g448","tokens":325,"max_abs_dp":9.747627500311662e-05,"mean_abs_dp":2.3997311036856045e-05,"calls":{"main":21,"prefill":19,"total":40},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g256","tokens":171,"max_abs_dp":8.184630676866256e-06,"mean_abs_dp":2.7938934091106095e-06,"calls":{"main":27,"prefill":9,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g448","tokens":303,"max_abs_dp":9.989629081941231e-06,"mean_abs_dp":4.426813022127694e-06,"calls":{"main":15,"prefill":18,"total":33},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g256","tokens":211,"max_abs_dp":0.0016373179818902558,"mean_abs_dp":0.00020212503903476042,"calls":{"main":35,"prefill":11,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g448","tokens":343,"max_abs_dp":0.0015474438071902963,"mean_abs_dp":0.00022706671983379652,"calls":{"main":23,"prefill":20,"total":43},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g256","tokens":177,"max_abs_dp":0.000224868824231228,"mean_abs_dp":5.478847091343327e-05,"calls":{"main":33,"prefill":9,"total":42},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g448","tokens":309,"max_abs_dp":7.153999954923052e-05,"mean_abs_dp":3.159614577444473e-05,"calls":{"main":21,"prefill":18,"total":39},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.053189992904663086},{"id":"r20","arm":"g256","tokens":143,"max_abs_dp":0.0010215486531204299,"mean_abs_dp":0.00020431019919575183,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r20","arm":"g448","tokens":275,"max_abs_dp":0.003167959185348068,"mean_abs_dp":0.0006335872444851548,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g256","tokens":143,"max_abs_dp":0.0003779720920398244,"mean_abs_dp":7.559234621520785e-05,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g448","tokens":275,"max_abs_dp":0.0003510978190582881,"mean_abs_dp":7.02183181676988e-05,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g256","tokens":147,"max_abs_dp":0.00010634076824134198,"mean_abs_dp":2.900891864273134e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g448","tokens":279,"max_abs_dp":0.00016255390723463048,"mean_abs_dp":3.946475624395112e-05,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.04187512397766113},{"id":"r23","arm":"g256","tokens":150,"max_abs_dp":7.761231113345701e-07,"mean_abs_dp":5.237824768687645e-07,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r23","arm":"g448","tokens":282,"max_abs_dp":6.35697471107477e-07,"mean_abs_dp":4.0557749720480154e-07,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g256","tokens":150,"max_abs_dp":3.1340847057026444e-05,"mean_abs_dp":2.089802593132409e-05,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g448","tokens":282,"max_abs_dp":2.1980285721440434e-05,"mean_abs_dp":1.464090366585422e-05,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g256","tokens":150,"max_abs_dp":0.00012555980040607206,"mean_abs_dp":8.37017993809521e-05,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g448","tokens":282,"max_abs_dp":0.00013002487058610956,"mean_abs_dp":8.668817530604759e-05,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g256","tokens":145,"max_abs_dp":0.0002712236654069614,"mean_abs_dp":0.00015280369152171437,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g448","tokens":277,"max_abs_dp":0.009645075075595422,"mean_abs_dp":0.00482252682758809,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g256","tokens":145,"max_abs_dp":0.0010564056541463307,"mean_abs_dp":0.0005281996838594962,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g448","tokens":277,"max_abs_dp":0.0018652498931602812,"mean_abs_dp":0.0009326308837615304,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g256","tokens":170,"max_abs_dp":0.007688615102949714,"mean_abs_dp":0.0025679425313154367,"calls":{"main":10,"prefill":10,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g448","tokens":302,"max_abs_dp":0.0032827771096515523,"mean_abs_dp":0.0010942566015300634,"calls":{"main":14,"prefill":18,"total":32},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g256","tokens":164,"max_abs_dp":0.009505770333302022,"mean_abs_dp":0.004752886097973589,"calls":{"main":4,"prefill":10,"total":14},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g448","tokens":296,"max_abs_dp":0.0010645366053126937,"mean_abs_dp":0.0005411707111177151,"calls":{"main":8,"prefill":18,"total":26},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g256","tokens":121,"max_abs_dp":2.1228588513722e-06,"mean_abs_dp":1.1953254186007538e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g448","tokens":253,"max_abs_dp":2.2734937715762626e-06,"mean_abs_dp":1.1509932108351578e-06,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g256","tokens":114,"max_abs_dp":0.000413855704184396,"mean_abs_dp":0.00020693592112921386,"calls":{"main":2,"prefill":7,"total":9},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g448","tokens":246,"max_abs_dp":0.001299537241839177,"mean_abs_dp":0.0006497690429251551,"calls":{"main":6,"prefill":15,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g256","tokens":105,"max_abs_dp":0.00025746877992327555,"mean_abs_dp":0.00025744992064115953,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g448","tokens":237,"max_abs_dp":0.0001315327521483889,"mean_abs_dp":0.00013152762987424216,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g256","tokens":144,"max_abs_dp":7.141881275282008e-05,"mean_abs_dp":2.5243524505169466e-05,"calls":{"main":16,"prefill":8,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g448","tokens":276,"max_abs_dp":7.615033003543353e-06,"mean_abs_dp":3.5500831285898546e-06,"calls":{"main":20,"prefill":16,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g256","tokens":121,"max_abs_dp":2.3593891182782656e-05,"mean_abs_dp":6.737934251871467e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g448","tokens":253,"max_abs_dp":4.6292827779126355e-05,"mean_abs_dp":1.3228457627311453e-05,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g256","tokens":145,"max_abs_dp":0.00024877619685478684,"mean_abs_dp":8.392341152053523e-05,"calls":{"main":17,"prefill":8,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g448","tokens":277,"max_abs_dp":0.0005070667012941135,"mean_abs_dp":0.00016933041561286444,"calls":{"main":21,"prefill":16,"total":37},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"t01","arm":"text","tokens":62,"max_abs_dp":1.7319149868410557e-06,"mean_abs_dp":1.1628657777583596e-06,"calls":{"main":14,"prefill":3,"total":17},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t02","arm":"text","tokens":107,"max_abs_dp":0.00014116063985836202,"mean_abs_dp":5.669456867849391e-05,"calls":{"main":27,"prefill":5,"total":32},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t03","arm":"text","tokens":51,"max_abs_dp":0.0003660741231484854,"mean_abs_dp":0.0001464325829255167,"calls":{"main":3,"prefill":3,"total":6},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t04","arm":"text","tokens":74,"max_abs_dp":0.0019238409545840884,"mean_abs_dp":0.000834853733393041,"calls":{"main":26,"prefill":3,"total":29},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t05","arm":"text","tokens":78,"max_abs_dp":2.9217003588018447e-06,"mean_abs_dp":1.4060929574488099e-06,"calls":{"main":14,"prefill":4,"total":18},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t06","arm":"text","tokens":95,"max_abs_dp":5.967880864876651e-05,"mean_abs_dp":2.2496402550165323e-05,"calls":{"main":15,"prefill":5,"total":20},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null}],"slots":[{"id":"r01","arm":"g256","t":104,"nopts":3,"probs":[0.999755396409714,0.0001365687873680299,0.00010803480291798109],"probs_oracle":[0.9997528195381165,0.00013744122406933457,0.0001098079519579187],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.5768715975482692e-06,"mean_abs_dp":1.7408191129301828e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996153783140471,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r01","arm":"g448","t":236,"nopts":3,"probs":[0.9996696973594793,0.0001809087391935138,0.00014939390132710846],"probs_oracle":[0.9996669292449951,0.00018169928807765245,0.00015134402201510966],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.768114484186235e-06,"mean_abs_dp":1.8362613521086973e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9994852299569175,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r02","arm":"g256","t":104,"nopts":3,"probs":[0.00019560398262911415,0.9996441238576415,0.00016027215972932892],"probs_oracle":[0.00019598512153606862,0.9996423721313477,0.00016159516235347837],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.7517262937971623e-06,"mean_abs_dp":1.151955941633693e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9994463870098116,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r02","arm":"g448","t":236,"nopts":3,"probs":[0.0002673149413439089,0.9994795963589859,0.00025308869967015395],"probs_oracle":[0.000269706710241735,0.9994760155677795,0.00025428077788092196],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.5807912063257064e-06,"mean_abs_dp":2.3882127716399297e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992063088575378,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r03","arm":"g256","t":104,"nopts":3,"probs":[0.0002693968925538489,0.0003052666720584039,0.9994253364353877],"probs_oracle":[0.00026771891862154007,0.0002989834756590426,0.9994333386421204],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.002206732693473e-06,"mean_abs_dp":5.321125688121196e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991343551664613,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r03","arm":"g448","t":236,"nopts":3,"probs":[0.0004440067556332185,0.0004763497139775458,0.9990796435303891],"probs_oracle":[0.00043719500536099076,0.000467298727016896,0.9990954399108887],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5796380499533136e-05,"mean_abs_dp":1.0553039244136881e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9986281411838718,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g256","t":112,"nopts":4,"probs":[0.9992562309491049,0.00043379801050544026,0.0002321953427918379,7.777569759792202e-05],"probs_oracle":[0.9992350339889526,0.0004433949070516974,0.0002398906071903184,8.168041676981375e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.11969601522366e-05,"mean_abs_dp":1.0598460067216487e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9987916390819009,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g256","t":152,"nopts":5,"probs":[0.9963882543759199,0.00140724936659051,0.0009085884798557473,0.0008208412871331412,0.00047506649050058314],"probs_oracle":[0.996371865272522,0.0014012253377586603,0.0009149523684754968,0.0008468246669508517,0.0004651718772947788],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.5983379817710525e-05,"mean_abs_dp":1.2931002774615302e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9949706399347633,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g448","t":244,"nopts":4,"probs":[0.9994403775955385,0.0003100785892044343,0.00016990904745579075,7.963476780132266e-05],"probs_oracle":[0.9994351267814636,0.00030979173607192934,0.0001716899569146335,8.336015162058175e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.250814074875265e-06,"mean_abs_dp":2.7609901213705197e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991253350453917,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g448","t":284,"nopts":5,"probs":[0.9983706657884381,0.0006257055511973763,0.00047601168152259414,0.0003220856148865449,0.00020553136395529638],"probs_oracle":[0.9983682036399841,0.0006189235136844218,0.0004789685772266239,0.0003326288133393973,0.00020131732162553817],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0543198452852415e-05,"mean_abs_dp":5.391664490720526e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9977492801262997,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r05","arm":"g256","t":104,"nopts":3,"probs":[0.005505057037375104,0.9932472550968242,0.0012476878658007774],"probs_oracle":[0.005630022846162319,0.9931067824363708,0.0012632559519261122],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00014047266045336126,"mean_abs_dp":9.366885178863718e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9874767595902085,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r05","arm":"g448","t":236,"nopts":3,"probs":[0.005458206369994767,0.9925180830803724,0.0020237105496327197],"probs_oracle":[0.005602917168289423,0.9923451542854309,0.0020519329700618982],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00017292879494146263,"mean_abs_dp":0.00011528733788843245,"finite":true,"read_from":"main","oracle_top2_margin":0.9867422371171415,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r06","arm":"g256","t":125,"nopts":8,"probs":[0.0004400236403511958,0.00041987349080713975,0.9978826403507972,0.00014064000275697136,0.00019073612980814877,0.00023187616405674332,0.0005921165281113695,0.00010209369331129326],"probs_oracle":[0.00044743184116669,0.00042624998604878783,0.997841477394104,0.0001445232192054391,0.00019250228069722652,0.00023652952222619206,0.0006083882763050497,0.00010282994480803609],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.116295669320813e-05,"mean_abs_dp":1.0282297243470957e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.997233089117799,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r06","arm":"g448","t":257,"nopts":8,"probs":[0.0004757428233225181,0.0004434410713410775,0.9978067717693796,0.00015324913356830282,0.00016441232034512368,0.0002336770236383557,0.000587463972707059,0.0001352418856978002],"probs_oracle":[0.00048085491289384663,0.0004459824995137751,0.9977728724479675,0.0001569857558934018,0.00016541295917704701,0.00023827303084544837,0.0006033360259607434,0.00013632865739054978],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.389932141206753e-05,"mean_abs_dp":8.480616558330334e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9971695364220068,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":112,"nopts":4,"probs":[0.9987818028366288,0.0005744161411319111,0.00024899096687484233,0.00039479005536444453],"probs_oracle":[0.9987712502479553,0.00058116385480389,0.0002529850753489882,0.000394599192077294],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0552588673440333e-05,"mean_abs_dp":5.37131852667891e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981900863931514,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":156,"nopts":6,"probs":[0.0016061032410859485,0.9957521667623974,0.0005464491065163368,0.0005507349599771212,0.000989491502219736,0.000555054427803211],"probs_oracle":[0.0015793738421052694,0.9958256483078003,0.0005403549876064062,0.0005451973993331194,0.0009696456836536527,0.0005397982313297689],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.348154540287588e-05,"mean_abs_dp":2.4490773162835463e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.994246274465695,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":183,"nopts":2,"probs":[0.06008665017400762,0.9399133498259924],"probs_oracle":[0.06064261496067047,0.9393573999404907],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.000555964786662852,"mean_abs_dp":0.000555957336082262,"finite":true,"read_from":"main","oracle_top2_margin":0.8787147849798203,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":244,"nopts":4,"probs":[0.9985792091788953,0.0005833435292802529,0.0002979427210720479,0.0005395045707521943],"probs_oracle":[0.9985532164573669,0.0005945708835497499,0.0003056767163798213,0.0005464947316795588],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.599272152836729e-05,"mean_abs_dp":1.2986058008250507e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9979586455738172,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":288,"nopts":6,"probs":[0.0017091046515143318,0.9954124140695666,0.0006336757680525778,0.0006094000306012636,0.0010447547175155257,0.0005906507627496457],"probs_oracle":[0.0016743798041716218,0.9955074787139893,0.0006255566840991378,0.0006036655395291746,0.001013218774460256,0.0005756971659138799],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.506464442265994e-05,"mean_abs_dp":3.16887677803224e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9938330989098176,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":315,"nopts":2,"probs":[0.07107366938018356,0.9289263306198166],"probs_oracle":[0.07187289744615555,0.9281271696090698],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0007992280659719908,"mean_abs_dp":0.0007991945383593602,"finite":true,"read_from":"main","oracle_top2_margin":0.8562542721629143,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r08","arm":"g256","t":119,"nopts":6,"probs":[0.0006816074483659527,0.0005391958716979997,0.9980078322636227,0.0003853459118428981,0.00013526913726709207,0.0002507493672031594],"probs_oracle":[0.0006678571808151901,0.0005291247507557273,0.9980358481407166,0.0003851840447168797,0.00013258290709927678,0.00024942721938714385],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.8015877093889863e-05,"mean_abs_dp":9.334585116129031e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9973679909599014,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r08","arm":"g448","t":251,"nopts":6,"probs":[0.0005697293907046283,0.0004911374340030258,0.9984022148036586,0.00026084112224639345,0.00010095680427838349,0.00017512044510889345],"probs_oracle":[0.0005603667814284563,0.00047818213352002203,0.9984318614006042,0.00025924580404534936,9.779584797797725e-05,0.00017250367091037333],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.9646596945598347e-05,"mean_abs_dp":9.889592567457425e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9978714946191758,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r09","arm":"g256","t":127,"nopts":8,"probs":[0.0006934185961940363,0.0006934185961940363,0.0005571774950576876,0.017064556152041468,0.9764060871361167,0.00349406590113812,0.0007675444182639585,0.0003237317049940974],"probs_oracle":[0.0006885188049636781,0.0006795556400902569,0.0005544525920413435,0.01720018871128559,0.9762108325958252,0.0035912375897169113,0.0007585337734781206,0.0003167196991853416],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00019525454029145628,"mean_abs_dp":5.81961361324308e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9590106438845396,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r09","arm":"g448","t":259,"nopts":8,"probs":[0.001144253149969525,0.0009266444402602349,0.0006674369523453798,0.002137746321709652,0.9546212389830444,0.037304992072413175,0.0024413829301876983,0.0007563051500698042],"probs_oracle":[0.0011424915865063667,0.0009199993801303208,0.0006675715558230877,0.0021507099736481905,0.9533334374427795,0.03857970982789993,0.0024612408597022295,0.0007448457181453705],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0012878015402648657,"mean_abs_dp":0.00032691769202498845,"finite":true,"read_from":"main","oracle_top2_margin":0.9147537276148796,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r10","arm":"g256","t":111,"nopts":3,"probs":[0.9985641439765004,0.0011332259530196082,0.00030263007048000824],"probs_oracle":[0.9985672235488892,0.0011292636627331376,0.00030339331715367734],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.962290286470575e-06,"mean_abs_dp":2.6017031163013717e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.997437959886156,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r10","arm":"g448","t":243,"nopts":3,"probs":[0.9995558538495318,0.00033270420206987523,0.00011144194839834729],"probs_oracle":[0.9995611310005188,0.0003284418198745698,0.00011045892460970208],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.277150986970902e-06,"mean_abs_dp":3.5075189903071883e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992326891806442,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r11","arm":"g256","t":102,"nopts":2,"probs":[0.002822832897761561,0.9971771671022386],"probs_oracle":[0.002824368653818965,0.9971755743026733],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5927995652420535e-06,"mean_abs_dp":1.564277811323114e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9943512056488544,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r11","arm":"g448","t":234,"nopts":2,"probs":[0.007345826574477061,0.9926541734255229],"probs_oracle":[0.007302414625883102,0.992697536945343],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.3411948593958735e-05,"mean_abs_dp":4.3387734207054295e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9853951223194599,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r12","arm":"g256","t":116,"nopts":5,"probs":[0.9988496853864509,0.0006509430002991509,0.00030036084881837144,0.00011625264899144839,8.27581154400108e-05],"probs_oracle":[0.9988265633583069,0.0006655653123743832,0.0003061715397052467,0.00011903217091457918,8.272226841654629e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.3122028144029194e-05,"mean_abs_dp":9.274080010546421e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981609980459325,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r12","arm":"g448","t":248,"nopts":5,"probs":[0.9991694490902987,0.00045103938414601543,0.00020171728985908093,0.00010182680826733488,7.596742742897676e-05],"probs_oracle":[0.9991558790206909,0.0004565727722365409,0.00020724281785078347,0.00010403238411527127,7.625338912475854e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.3570069607782287e-05,"mean_abs_dp":5.424104646745695e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986993062484544,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r13","arm":"g256","t":139,"nopts":9,"probs":[6.699639880964326e-05,0.000125165751238633,9.672059604174703e-05,4.0161953568592486e-05,0.00012419170363479095,0.00017377543614974488,0.00020961318769014648,0.0161393209562696,0.983024054016597],"probs_oracle":[6.637163460254669e-05,0.00012154777505202219,9.527682414045557e-05,4.0484348573954776e-05,0.00012243450328242034,0.00017381930956616998,0.00021287811978254467,0.017157258465886116,0.9820099472999573],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010179375096165157,"mean_abs_dp":0.0002270132377130864,"finite":true,"read_from":"main","oracle_top2_margin":0.9648526888340712,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r13","arm":"g448","t":271,"nopts":9,"probs":[2.045735459059768e-05,3.3728475655773307e-05,3.3465998575832204e-05,9.042504708183235e-06,2.7314151982633743e-05,3.193347892437445e-05,4.399022523624303e-05,9.312730755592179e-05,0.9997069405027705],"probs_oracle":[2.0277637304388918e-05,3.2944568374659866e-05,3.3252264984184876e-05,9.089779268833809e-06,2.7153426344739273e-05,3.251702946727164e-05,4.40161456936039e-05,9.227909322362393e-05,0.9997084736824036],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5331796330286807e-06,"mean_abs_dp":4.862470358999081e-07,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9996161945891799,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r14","arm":"g256","t":110,"nopts":4,"probs":[0.9989656216236307,0.0006260784259036891,0.00017797856901463773,0.00023032138145094542],"probs_oracle":[0.9989142417907715,0.0006516185821965337,0.0001856259477790445,0.00024845064035616815],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.137983285918857e-05,"mean_abs_dp":2.5674156705415657e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.998262623208575,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r14","arm":"g448","t":242,"nopts":4,"probs":[0.9992771739621324,0.0004475776179294981,0.00011813389925229162,0.00015711452068568034],"probs_oracle":[0.9992570281028748,0.0004573110491037369,0.00012069327203789726,0.00016500464698765427],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.0145859257691967e-05,"mean_abs_dp":1.0082197379877578e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.998799717053771,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g256","t":114,"nopts":4,"probs":[0.0013021585212991049,0.0009526795255484737,0.9968978312427982,0.0008473307103542585],"probs_oracle":[0.0013125522527843714,0.0009568282985128462,0.9968631267547607,0.00086748949252069],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.4704488037440306e-05,"mean_abs_dp":1.7351443663377714e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9955505745019764,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g256","t":146,"nopts":3,"probs":[0.9973152493968619,0.0019709278426206776,0.0007138227605173603],"probs_oracle":[0.9973812699317932,0.001912115840241313,0.000706570572219789],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.602053493132143e-05,"mean_abs_dp":4.402824186941909e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9954691540915519,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g448","t":246,"nopts":4,"probs":[0.006571974705538923,0.0033045964115712466,0.9830174299221682,0.007105998960721564],"probs_oracle":[0.006657375022768974,0.0033173311967402697,0.9827334880828857,0.007291808724403381],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0002839418392824866,"mean_abs_dp":0.00014197167634084473,"finite":true,"read_from":"main","oracle_top2_margin":0.9754416793584824,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g448","t":278,"nopts":3,"probs":[0.9946000492337841,0.004464202985831848,0.0009357477803842864],"probs_oracle":[0.9947662353515625,0.004311799071729183,0.0009220570791512728],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00016618611777841075,"mean_abs_dp":0.00011076024437136306,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9904544362798333,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":117,"nopts":4,"probs":[0.0007721859177778627,0.0010889566850533462,0.997779757596133,0.0003590998010356395],"probs_oracle":[0.0007772505050525069,0.0010821042815223336,0.997784435749054,0.0003561374032869935],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.852403531012541e-06,"mean_abs_dp":4.889385368809157e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9967023314675316,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":165,"nopts":7,"probs":[0.0015807213866532268,0.003851668059482278,0.9877022803627443,0.0037919534810878382,0.0010779583298369502,0.0010366623947264762,0.0009587559854686503],"probs_oracle":[0.001566192484460771,0.0037890495732426643,0.9877481460571289,0.003863656660541892,0.0010514376917853951,0.0010348198702558875,0.0009466473711654544],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.170317945405383e-05,"mean_abs_dp":3.3598291299435215e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.983884489396587,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":192,"nopts":2,"probs":[0.003621685478675127,0.996378314521325],"probs_oracle":[0.0035970157478004694,0.9964029788970947],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.4669730874657403e-05,"mean_abs_dp":2.466705332220681e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9928059631492943,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":249,"nopts":4,"probs":[0.00016987713906667826,0.0004304206405629625,0.9992526858101947,0.00014701641017563836],"probs_oracle":[0.00016730134666431695,0.00042254829895682633,0.9992651343345642,0.00014504071441479027],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2448524369546021e-05,"mean_abs_dp":6.218088534722902e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9988425860356074,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":297,"nopts":7,"probs":[0.001227386572953095,0.002618756721822501,0.9924728643583892,0.0017174225877960656,0.0006123658821189787,0.0006993437885740216,0.000651860088346],"probs_oracle":[0.0012208223342895508,0.0025618993677198887,0.9925703406333923,0.0017182007431983948,0.0005935655790381134,0.0006946290959604084,0.0006406526081264019],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.747627500311662e-05,"mean_abs_dp":2.805692844081124e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9900084412656724,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":324,"nopts":2,"probs":[0.0044683842664911206,0.9955316157335089],"probs_oracle":[0.004423052072525024,0.9955769777297974],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.536199628846216e-05,"mean_abs_dp":4.534709512727915e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9911539256572723,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":108,"nopts":3,"probs":[0.999793305333327,0.00014090928886091258,6.578537781193891e-05],"probs_oracle":[0.9997918009757996,0.00014244449266698211,6.58148928778246e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5352038060695327e-06,"mean_abs_dp":1.023025466464276e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996493564831326,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":141,"nopts":3,"probs":[0.9992366449824038,0.00039806762839613775,0.0003652873892001707],"probs_oracle":[0.999238133430481,0.00039801813545636833,0.0003638898197095841],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.4884480771248931e-06,"mean_abs_dp":9.785035024936258e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9988401152950246,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":170,"nopts":2,"probs":[0.9981613347123114,0.0018386652876885773],"probs_oracle":[0.9981531500816345,0.0018468272173777223],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.184630676866256e-06,"mean_abs_dp":8.173280183005585e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9963063228642568,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":240,"nopts":3,"probs":[0.9998114905071781,0.00012930798386508665,5.920150895686227e-05],"probs_oracle":[0.9998090863227844,0.0001311514643020928,5.978461922495626e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.404184393678399e-06,"mean_abs_dp":1.6102583662595095e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996779348584823,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":273,"nopts":3,"probs":[0.999184405694298,0.000423718631874856,0.0003918756738270545],"probs_oracle":[0.9991744160652161,0.00042735174065455794,0.0003981590270996094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.989629081941231e-06,"mean_abs_dp":6.6353637113993474e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9987470643245615,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":302,"nopts":2,"probs":[0.9979966667138236,0.0020033332861764124],"probs_oracle":[0.9979913234710693,0.002008667681366205],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.343242754252131e-06,"mean_abs_dp":5.338818972022491e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9959826557897031,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":123,"nopts":6,"probs":[0.9988440406451573,0.0006875289446565364,0.00016980766767819388,7.162077460333754e-05,0.0001252083224842479,0.00010179364542055033],"probs_oracle":[0.9988755583763123,0.000659995072055608,0.00016870087711140513,7.21013275324367e-05,0.00012259691720828414,0.00010095535981236026],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.151773115495082e-05,"mean_abs_dp":1.0681439689320165e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9982155633042566,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":164,"nopts":5,"probs":[0.003240176742589289,0.01607379616220138,0.948911348455058,0.028431647646852916,0.0033430309932984156],"probs_oracle":[0.0031590552534908056,0.015553153119981289,0.9480063915252686,0.03006896562874317,0.0032123548444360495],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0016373179818902558,"mean_abs_dp":0.0006549431183721239,"finite":true,"read_from":"main","oracle_top2_margin":0.9179374258965254,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":210,"nopts":6,"probs":[0.000534388468287204,0.0009233404731951682,0.0005470611256128158,0.000467926177544461,0.9968674020773639,0.0006598816779964693],"probs_oracle":[0.0005212401156313717,0.0009053464164026082,0.000543770962394774,0.0004741615557577461,0.9969097971916199,0.0006456233095377684],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.2395114255966426e-05,"mean_abs_dp":1.622023893239777e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9960044507752173,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":255,"nopts":6,"probs":[0.9989622667020183,0.00044743657053541845,0.0001985487746069953,8.539476386412036e-05,0.0001492881805814305,0.0001570650083936732],"probs_oracle":[0.9989650249481201,0.00044237051042728126,0.00019978830823674798,8.695968426764011e-05,0.0001475344179198146,0.0001582982367835939],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.066060108137186e-06,"mean_abs_dp":2.2692918824664924e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9985226544376928,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":296,"nopts":5,"probs":[0.008433652843579347,0.04559189652476152,0.9229199410080258,0.0181352034795276,0.004919306144105839],"probs_oracle":[0.008088547736406326,0.04423723742365837,0.9244673848152161,0.018492545932531357,0.004714324604719877],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0015474438071902963,"mean_abs_dp":0.0007619064015712366,"finite":true,"read_from":"main","oracle_top2_margin":0.8802301473915577,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":342,"nopts":6,"probs":[0.0006342687066817879,0.0008669425331919015,0.0005641302656092515,0.00048631068646836864,0.9963438343669072,0.0011045134411413573],"probs_oracle":[0.0006307050934992731,0.0008567753247916698,0.0005710572004318237,0.0004954338655807078,0.9963390827178955,0.0011069673346355557],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0167208400231696e-05,"mean_abs_dp":6.164413003926513e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.99523211538326,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":110,"nopts":3,"probs":[0.9930316933687084,0.005858830910290339,0.0011094757210011975],"probs_oracle":[0.9930190443992615,0.005865252576768398,0.0011157557601109147],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2648969446971137e-05,"mean_abs_dp":8.450225011582407e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9871537918224931,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":146,"nopts":4,"probs":[0.002499652361071137,0.985071440170062,0.011558275415834056,0.0008706320530326969],"probs_oracle":[0.0024147704243659973,0.9852963089942932,0.011448029428720474,0.0008408919093199074],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.000224868824231228,"mean_abs_dp":0.00011243422294068484,"finite":true,"read_from":"main","oracle_top2_margin":0.9738482795655727,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":176,"nopts":2,"probs":[0.9979809860995514,0.0020190139004485808],"probs_oracle":[0.9979719519615173,0.0020279884338378906],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.034138034103023e-06,"mean_abs_dp":9.004335711706437e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9959439635276794,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g448","t":242,"nopts":3,"probs":[0.9957976485505746,0.0035634585384851643,0.0006388929109403547],"probs_oracle":[0.9957261085510254,0.0036251493729650974,0.0006488442304544151],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.153999954923052e-05,"mean_abs_dp":4.772738451440803e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9921009591780603,"python":{"ref":"e2e","max_abs_dp":5.007672684093122e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":278,"nopts":4,"probs":[0.0007311464943971777,0.9978555869855932,0.0011867679022261688,0.00022649861778363273],"probs_oracle":[0.000710506399627775,0.9978955984115601,0.0011723366333171725,0.00022145592083688825],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.0011425966879344e-05,"mean_abs_dp":2.0031371648005717e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9967232617782429,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":308,"nopts":2,"probs":[0.998175617214723,0.0018243827852769423],"probs_oracle":[0.9981451034545898,0.0018549266969785094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.054391170156717e-05,"mean_abs_dp":3.052883591737779e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9962901767576113,"python":{"ref":"e2e","max_abs_dp":2.8676610337496954e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r20","arm":"g256","t":142,"nopts":10,"probs":[0.0003752092940643742,0.0010199246059241348,0.0004285024637740692,0.0005676735330856547,0.0006688828868548242,0.00033897334900505897,0.0014160250825835967,0.46991001024293366,0.0010523004883083464,0.5242224980534663],"probs_oracle":[0.0003781993000302464,0.0010252433130517602,0.00043770880438387394,0.0005875966744497418,0.0006762796547263861,0.00034810963552445173,0.0014829656574875116,0.46888846158981323,0.001062334282323718,0.5251131057739258],"probs_swift_json_equal":true,"argmax":9,"argmax_oracle":9,"argmax_equal":true,"full_vocab_top1_id":41,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010215486531204299,"mean_abs_dp":0.00020431019919575183,"finite":true,"read_from":"main","oracle_top2_margin":0.05622464418411255,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r20","arm":"g448","t":274,"nopts":10,"probs":[0.0004462873107688687,0.0010705876038144555,0.00061478673737031,0.0006595698889443333,0.0009596685257546854,0.0004127482897180398,0.002000120962903406,0.8931639513690334,0.0012419033471868894,0.09943037596450577],"probs_oracle":[0.0004485415993258357,0.0010800594463944435,0.0006290642777457833,0.0006908925715833902,0.000972151174210012,0.00042305333772674203,0.0021294124890118837,0.8899959921836853,0.0012687055859714746,0.10236208140850067],"probs_swift_json_equal":true,"argmax":7,"argmax_oracle":7,"argmax_equal":true,"full_vocab_top1_id":39,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.003167959185348068,"mean_abs_dp":0.0006335872444851548,"finite":true,"read_from":"main","oracle_top2_margin":0.7876339107751846,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r21","arm":"g256","t":142,"nopts":10,"probs":[0.0003806633017063551,0.0006839280772388108,0.0003836488809845655,0.970595676864989,0.00092754402703674,0.0004849776279125633,0.0005162559976120511,0.0007280376485190599,0.00042466060008941175,0.024874606973911684],"probs_oracle":[0.00038343475898727775,0.000686735613271594,0.0003942575422115624,0.9702177047729492,0.0009427570621483028,0.0004960547666996717,0.0005290430854074657,0.0007387973600998521,0.0004333727993071079,0.02517782151699066],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003779720920398244,"mean_abs_dp":7.559234621520785e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9450398832559586,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r21","arm":"g448","t":274,"nopts":10,"probs":[0.0005420045756578569,0.0008797602644334618,0.0005336015735757884,0.9647742774455231,0.0012119209522387742,0.0006692863152777638,0.0006692863152777638,0.001004717758338862,0.0005814860276640014,0.02913365877201275],"probs_oracle":[0.0005481093539856374,0.0008933480712585151,0.0005559168639592826,0.9644231796264648,0.0012373479548841715,0.0006897316197864711,0.0006905003683641553,0.0010131854796782136,0.0005959380650892854,0.02935273014008999],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003510978190582881,"mean_abs_dp":7.02183181676988e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9350704494863749,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r22","arm":"g256","t":146,"nopts":9,"probs":[0.001497815420134219,0.0019535195828734113,0.0009822913344533718,0.9885064424299235,0.0017511236366360415,0.002112258300986467,0.0015095629320716485,0.0008401983029792642,0.0008467880599422232],"probs_oracle":[0.0015125663485378027,0.0019855706486850977,0.0009997023735195398,0.9884001016616821,0.0017508834134787321,0.0021429003681987524,0.0015369588509202003,0.0008162282756529748,0.0008550662896595895],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00010634076824134198,"mean_abs_dp":2.900891864273134e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9862572012934834,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r22","arm":"g448","t":278,"nopts":9,"probs":[0.0024617606804946315,0.0036667815582753717,0.0015770630966120535,0.9854164243944478,0.0015048441194412314,0.0016919417226531283,0.0017870466083147272,0.0006943714839040915,0.0011997663358571152],"probs_oracle":[0.0024558305740356445,0.0037613804452121258,0.0016030978877097368,0.9852538704872131,0.0015132761327549815,0.0017176602268591523,0.0018075669649988413,0.0006852896767668426,0.0012020787689834833],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00016255390723463048,"mean_abs_dp":3.946475624395112e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.981492490042001,"python":{"ref":"e2e","max_abs_dp":1.911077835882501e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r23","arm":"g256","t":149,"nopts":3,"probs":[0.9997602713814426,1.8054515155315643e-05,0.00022167410340184044],"probs_oracle":[0.9997594952583313,1.821711521188263e-05,0.00022230672766454518],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.761231113345701e-07,"mean_abs_dp":5.237824768687645e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9995371885306668,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r23","arm":"g448","t":281,"nopts":3,"probs":[0.9997159043093791,2.2911355749567803e-05,0.00026118433487140033],"probs_oracle":[0.999715268611908,2.3231894374475814e-05,0.00026144483126699924],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.35697471107477e-07,"mean_abs_dp":4.0557749720480154e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.999453823780641,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r24","arm":"g256","t":149,"nopts":3,"probs":[3.92579615874165e-05,0.9991752989387929,0.0007854430996196572],"probs_oracle":[3.9324964745901525e-05,0.9991439580917358,0.000816729327198118],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.1340847057026444e-05,"mean_abs_dp":2.089802593132409e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9983272287645377,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r24","arm":"g448","t":281,"nopts":3,"probs":[3.8044285043092655e-05,0.9990220475937871,0.0009399081211699353],"probs_oracle":[3.7778951082145795e-05,0.9990003705024719,0.0009618884068913758],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.1980285721440434e-05,"mean_abs_dp":1.464090366585422e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9980384820955805,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r25","arm":"g256","t":149,"nopts":3,"probs":[0.0012238810106976132,0.001376046408269225,0.9974000725810331],"probs_oracle":[0.001173934550024569,0.0013004472712054849,0.9975256323814392],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012555980040607206,"mean_abs_dp":8.37017993809521e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9962251851102337,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r25","arm":"g448","t":281,"nopts":3,"probs":[0.002628406563880659,0.0012415713477821624,0.9961300220883372],"probs_oracle":[0.0025366705376654863,0.0012032677186653018,0.9962600469589233],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00013002487058610956,"mean_abs_dp":8.668817530604759e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9937233764212579,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r26","arm":"g256","t":144,"nopts":4,"probs":[0.986467419916574,0.0013088256685909964,0.010873382159700651,0.001350372255134366],"probs_oracle":[0.986732542514801,0.001274453941732645,0.01060215849429369,0.0013908690307289362],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0002712236654069614,"mean_abs_dp":0.00015280369152171437,"finite":true,"read_from":"main","oracle_top2_margin":0.9761303840205073,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r26","arm":"g448","t":276,"nopts":4,"probs":[0.8428485100197483,0.02975619181099293,0.0854323417771534,0.04196295639210546],"probs_oracle":[0.8332034349441528,0.030570058152079582,0.08590248227119446,0.050323981791734695],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.009645075075595422,"mean_abs_dp":0.00482252682758809,"finite":true,"read_from":"main","oracle_top2_margin":0.7473009526729584,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r27","arm":"g256","t":144,"nopts":4,"probs":[0.0078269701772855,0.008731616201164524,0.005254779190857521,0.9781866344306925],"probs_oracle":[0.007379894610494375,0.008370750583708286,0.005006327293813229,0.9792430400848389],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010564056541463307,"mean_abs_dp":0.0005281996838594962,"finite":true,"read_from":"main","oracle_top2_margin":0.9708722895011306,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r27","arm":"g448","t":276,"nopts":4,"probs":[0.01099636501398774,0.013900707872412706,0.008171804593124487,0.966931122520475],"probs_oracle":[0.010312587022781372,0.013162228278815746,0.007728788536041975,0.9687963724136353],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0018652498931602812,"mean_abs_dp":0.0009326308837615304,"finite":true,"read_from":"main","oracle_top2_margin":0.9556341441348195,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r28","arm":"g256","t":169,"nopts":6,"probs":[0.027973158830786167,0.7985582683278948,0.17002286656753948,0.0012386944192324437,0.0008251483729857597,0.00138186348156154],"probs_oracle":[0.029139939695596695,0.7908696532249451,0.17654110491275787,0.0012475387193262577,0.0008099511615000665,0.0013918428448960185],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.007688615102949714,"mean_abs_dp":0.0025679425313154367,"finite":true,"read_from":"main","oracle_top2_margin":0.6143285483121872,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r28","arm":"g448","t":301,"nopts":6,"probs":[0.024059107351089543,0.7426323536800727,0.23005677100131036,0.0012455486416929082,0.0007703636205712258,0.0012358557052633843],"probs_oracle":[0.023014351725578308,0.7404738664627075,0.23333954811096191,0.0012221261858940125,0.000736995309125632,0.001213126815855503],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0032827771096515523,"mean_abs_dp":0.0010942566015300634,"finite":true,"read_from":"main","oracle_top2_margin":0.5071343183517456,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r29","arm":"g256","t":163,"nopts":4,"probs":[0.007622201780028129,0.8810067392664075,0.05416325765231141,0.057207801301252956],"probs_oracle":[0.007844753563404083,0.8715009689331055,0.05929643288254738,0.06135784834623337],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.009505770333302022,"mean_abs_dp":0.004752886097973589,"finite":true,"read_from":"main","oracle_top2_margin":0.8101431205868721,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r29","arm":"g448","t":295,"nopts":4,"probs":[0.002484101507144741,0.9789431071619533,0.015823141614734516,0.002749649716167261],"probs_oracle":[0.002556135645136237,0.9778785705566406,0.016833467409014702,0.002731863409280777],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010645366053126937,"mean_abs_dp":0.0005411707111177151,"finite":true,"read_from":"main","oracle_top2_margin":0.9610451031476259,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r30","arm":"g256","t":120,"nopts":4,"probs":[0.9998231301105909,6.761103250243447e-05,6.036985690052259e-05,4.888900000639292e-05],"probs_oracle":[0.9998210072517395,6.738237425452098e-05,6.102487168391235e-05,5.0663769798120484e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.1228588513722e-06,"mean_abs_dp":1.1953254186007538e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999753624877485,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r30","arm":"g448","t":252,"nopts":4,"probs":[0.9998536195097018,5.327783637121997e-05,5.5617014284713345e-05,3.748563964216082e-05],"probs_oracle":[0.9998513460159302,5.3302719607017934e-05,5.6550619774498045e-05,3.8857629988342524e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.2734937715762626e-06,"mean_abs_dp":1.1509932108351578e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9997947953961557,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r31","arm":"g256","t":113,"nopts":4,"probs":[0.008902913131610247,0.0011230700387340725,0.00035895474438093475,0.9896150620852748],"probs_oracle":[0.00852695107460022,0.0010953464079648256,0.0003487524518277496,0.9900289177894592],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.000413855704184396,"mean_abs_dp":0.00020693592112921386,"finite":true,"read_from":"main","oracle_top2_margin":0.981501966714859,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r31","arm":"g448","t":245,"nopts":4,"probs":[0.05255451363294278,0.0007555294522778163,0.000469118904848662,0.9462208380099308],"probs_oracle":[0.05128991976380348,0.000733718799892813,0.00045598449651151896,0.94752037525177],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.001299537241839177,"mean_abs_dp":0.0006497690429251551,"finite":true,"read_from":"main","oracle_top2_margin":0.8962304554879665,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r32","arm":"g256","t":104,"nopts":2,"probs":[0.007999754508377905,0.9920002454916221],"probs_oracle":[0.007742323447018862,0.9922577142715454],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00025746877992327555,"mean_abs_dp":0.00025744992064115953,"finite":true,"read_from":"main","oracle_top2_margin":0.9845153908245265,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r32","arm":"g448","t":236,"nopts":2,"probs":[0.003794834220043924,0.9962051657799561],"probs_oracle":[0.0036633117124438286,0.9963366985321045],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0001315327521483889,"mean_abs_dp":0.00013152762987424216,"finite":true,"read_from":"main","oracle_top2_margin":0.9926733868196607,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g256","t":114,"nopts":4,"probs":[0.0008617448024981268,0.9981378520336721,0.0007313533150063774,0.00026904984882342793],"probs_oracle":[0.0008657005382701755,0.9981381893157959,0.000728585640899837,0.0002674604766070843],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.955735772048687e-06,"mean_abs_dp":2.1625160546796178e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9972724887775257,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g256","t":143,"nopts":2,"probs":[0.0044337652006436815,0.9955662347993564],"probs_oracle":[0.0045051840133965015,0.9954948425292969],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.141881275282008e-05,"mean_abs_dp":7.140554140614916e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9909896585159004,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g448","t":246,"nopts":4,"probs":[0.0005655814386908304,0.998906836993382,0.00038569303116014277,0.000141888536766913],"probs_oracle":[0.0005686153890565038,0.9989047050476074,0.00038564109127037227,0.00014100427506491542],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.033950365673361e-06,"mean_abs_dp":1.525524433006878e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9983360896585509,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g448","t":275,"nopts":2,"probs":[0.02017731832286999,0.97982268167713],"probs_oracle":[0.020169703289866447,0.979830265045166],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.615033003543353e-06,"mean_abs_dp":7.599200519755808e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9596605617552996,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r34","arm":"g256","t":120,"nopts":7,"probs":[0.9988502764587792,0.0005271500184296672,0.0001850470603471304,6.996211791949e-05,8.809560744718788e-05,7.303382375750886e-05,0.00020643491331983596],"probs_oracle":[0.9988266825675964,0.0005331173888407648,0.000190935141290538,7.275486859725788e-05,8.913177589420229e-05,7.561183883808553e-05,0.00021174417634028941],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.3593891182782656e-05,"mean_abs_dp":6.737934251871467e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9982935651787557,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r34","arm":"g448","t":252,"nopts":7,"probs":[0.9984354176980794,0.0006765928917309875,0.00024313874571634818,9.633685929236129e-05,0.00012913001311531475,9.978394018967788e-05,0.0003195998518760412],"probs_oracle":[0.9983891248703003,0.0006920418818481266,0.0002512198407202959,0.00010025671508628875,0.00013134843902662396,0.00010331271914765239,0.000332709081703797],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.6292827779126355e-05,"mean_abs_dp":1.3228457627311453e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9976970829884522,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g256","t":113,"nopts":4,"probs":[0.0002340860588611912,0.9995533624648736,0.00012335468463597186,8.919679162922821e-05],"probs_oracle":[0.00023222366871777922,0.9995563626289368,0.00012253291788510978,8.888279262464494e-05],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.0001640631827087e-06,"mean_abs_dp":1.4995799905100092e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999324138960219,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g256","t":144,"nopts":2,"probs":[0.986878728873449,0.013121271126550956],"probs_oracle":[0.9866299629211426,0.013370047323405743],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00024877619685478684,"mean_abs_dp":0.0002487710745805857,"finite":true,"read_from":"main","oracle_top2_margin":0.9732599155977368,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g448","t":245,"nopts":4,"probs":[0.00027152594743203213,0.9994847802664772,0.00013441509092575185,0.0001092786951649191],"probs_oracle":[0.0002710831176955253,0.9994843006134033,0.00013441758346743882,0.00011026520951418206],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.865143492629631e-07,"mean_abs_dp":4.778724253456796e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9992132174957078,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g448","t":276,"nopts":2,"probs":[0.9862576795040977,0.01374232049590237],"probs_oracle":[0.985750675201416,0.014249387197196484],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0005070667012941135,"mean_abs_dp":0.0005070355019879021,"finite":true,"read_from":"main","oracle_top2_margin":0.9715012880042195,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t01","arm":"text","t":61,"nopts":3,"probs":[0.0007730925428213612,0.9989512528219664,0.00027565463521236894],"probs_oracle":[0.0007738324929960072,0.9989522695541382,0.0002739227202255279],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.7319149868410557e-06,"mean_abs_dp":1.1628657777583596e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981784370611422,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":45,"nopts":2,"probs":[0.0009622125595974061,0.9990377874404025],"probs_oracle":[0.0009324414422735572,0.9990676045417786],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.981710137606175e-05,"mean_abs_dp":2.979410934995534e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.998135163099505,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":71,"nopts":2,"probs":[0.0037362205834114964,0.9962637794165885],"probs_oracle":[0.00368038984015584,0.9963196516036987],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.5872187110228566e-05,"mean_abs_dp":5.5851465182942504e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9926392617635429,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":106,"nopts":4,"probs":[0.9949236343773841,0.0028608062332513957,0.0012497936839403207,0.0009657657054241225],"probs_oracle":[0.9950647950172424,0.002778948051854968,0.001205416745506227,0.0009508960647508502],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00014116063985836202,"mean_abs_dp":7.05663500905389e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9922858469653875,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t03","arm":"text","t":50,"nopts":5,"probs":[0.004913862395458331,0.0021300016857536868,0.004616146523161474,0.9813559827425437,0.006984006653082918],"probs_oracle":[0.00478515587747097,0.0020656364504247904,0.004542955197393894,0.9817220568656921,0.006884180940687656],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003660741231484854,"mean_abs_dp":0.0001464325829255167,"finite":true,"read_from":"main","oracle_top2_margin":0.9748378759250045,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t04","arm":"text","t":46,"nopts":3,"probs":[0.01653248493810134,0.12215945866235682,0.8613080563995419],"probs_oracle":[0.016153892502188683,0.12061421573162079,0.863231897354126],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0019238409545840884,"mean_abs_dp":0.0012825587737442593,"finite":true,"read_from":"main","oracle_top2_margin":0.7426176816225052,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t04","arm":"text","t":73,"nopts":2,"probs":[0.9873750507184357,0.012624949281564409],"probs_oracle":[0.9872117638587952,0.012788254767656326],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00016330548609191746,"mean_abs_dp":0.0001632961728662138,"finite":true,"read_from":"main","oracle_top2_margin":0.9744235090911388,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t05","arm":"text","t":77,"nopts":7,"probs":[0.9990050369091737,0.0003172936226450292,0.00020486015990573076,8.776575141737641e-05,0.00011581743319913373,0.00014301544044474732,0.00012621068321426118],"probs_oracle":[0.9990049004554749,0.00031437192228622735,0.00020690326346084476,8.894488564692438e-05,0.00011396020272513852,0.00014418178761843592,0.00012674936442635953],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.9217003588018447e-06,"mean_abs_dp":1.4060929574488099e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986905285331886,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t06","arm":"text","t":63,"nopts":4,"probs":[0.9949011948265626,0.00042854626828138064,0.00023945546736189676,0.004430803437794217],"probs_oracle":[0.9948415160179138,0.0004282791051082313,0.00024067162303254008,0.004489557351917028],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.967880864876651e-05,"mean_abs_dp":2.99790104038426e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9903519586659968,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t06","arm":"text","t":94,"nopts":3,"probs":[0.0009771737282980766,0.9988420673074887,0.00018075896421321802],"probs_oracle":[0.0009630977874621749,0.9988608360290527,0.00017604485037736595],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.8768721564033086e-05,"mean_abs_dp":1.2519592078595622e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9978977382415906,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}}],"asset_identity":{"decoder_aimodelc_tree_sha256":"b75795b701f2f3c5fc5af4905819d59c73f2872fb0a92889713fed5295aedaa4","decoder_equals_round6":true,"tower_g256_equals_round6":true,"tower_g448_equals_round6":true},"vs_python_attribution":{"runs_tower_output_differs_from_python":["r19/g448","r22/g448"],"runs_whose_tile_differs_pillow_vs_resize_bicubic":["r19/g448","r22/g448"],"runs_with_slot_logits_differing_from_python":["r19/g448","r22/g448"],"same_set":true,"reading":"every Swift/Python difference sits on a run whose tile Pillow and resize_bicubic resize differently; the rest is bit-equal (tower output and fp16 logits)"}},"s3_alt":{"asset":"aot","bundle":"decider_2b_vision_decode_fp16_pf16","fixture_json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/aot_fp16_pf16/fixture_aot.json","label":"extra arm: AOT fp16_pf16 (the alternative ship candidate; correctness run, final binary, GPU shared with round 7)","assets":{"asset":"aot","bundle":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles/decider_2b_vision_decode_fp16_pf16","decoder":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles_aotc/decider_2b_vision_decode_fp16_pf16.h16c.aimodelc","decoder_options":"SpecializationOptions.default","tower_options":"SpecializationOptions.default"},"environment":{"argv":["swift/.build/release/decider-vision","fixture","--bundle","exports/bundles/decider_2b_vision_decode_fp16_pf16","--tower-g256","exports/decider_2b_vision_g256_vision_fp16w32_aotc/decider_2b_vision_g256_vision_fp16w32.h16c.aimodelc","--tower-g448","exports/decider_2b_vision_g448_vision_fp16w32_aotc/decider_2b_vision_g448_vision_fp16w32.h16c.aimodelc","--rows","fixtures/rows.json","--images","fixtures/images","--meta","fixtures/meta.json","--out","swift/aot_fp16_pf16/fixture_aot.json","--label","extra arm: AOT fp16_pf16 (the alternative ship candidate; correctness run, final binary, GPU shared with round 7)"],"build_configuration":"Release","chip":"Apple M4 Max","device_architecture":"h16c","model":"Mac16,9","os":"Version 27.0 (Build 26A428)","os_build":"26A428","pid":82198},"load_first":{"decoder_s":{"main":4.519789917,"model":0.001339375,"prefill":12.572404583},"tokenizer_s":0.7483882500000001,"tower_s":{"g256":0.9727998750000001,"g448":0.847584333},"wall_s":19.667627416},"function_names":["main","prefill"],"reset_check":{"bit_equal":true,"run":"r01/g256","slot_logits_max_abs_diff":0,"wall_from_file_s":0.37474866700000004},"bar":{"max_abs_dp":0.02,"mean_of_run_mean_abs_dp":0.002,"source":"readout_gate_vision.BAR (round 4, unchanged)","slots":"all","argmax":"all","full_vocab_top1":"the oracle's argmax letter on every slot","reset":"bit-equal re-run"},"summary":{"runs":76,"runs_expected":76,"slots_read":108,"slots_expected":108,"argmax_equal":108,"full_vocab_top1_is_oracle_letter":108,"max_abs_dp":0.00854564497937682,"worst":{"id":"r26","arm":"g448","t":276,"max_abs_dp":0.00854564497937682,"probs":[0.8417490799235297,0.028803069605559367,0.0826958537706501,0.04675199670026068],"probs_oracle":[0.8332034349441528,0.030570058152079582,0.08590248227119446,0.050323981791734695]},"mean_of_run_mean_abs_dp":0.00016557540518108565,"finite_all":true,"reset_bit_equal":true,"swift_json_probs_equal_recomputed":108,"slots_read_from_prefill":9},"checks":{"all_runs":true,"slots":true,"argmax":true,"full_vocab_top1":true,"max_abs_dp":true,"mean_of_run_mean_abs_dp":true,"finite":true,"reset_bit_equal":true},"result":"PASS","per_arm":{"g256":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.002927743999633692,"mean_of_run_mean_abs_dp":9.064051534252789e-05},"g448":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.00854564497937682,"mean_of_run_mean_abs_dp":0.0002544622062187517},"text":{"runs":6,"slots":10,"argmax_equal":10,"full_vocab_top1_is_oracle_letter":10,"max_abs_dp":0.0007736432970376628,"mean_of_run_mean_abs_dp":8.418925651962004e-05}},"vs_python_same_bundle":{"definition":"per slot, the same bundle read by Python: image runs vs the round-6 e2e arm (decider_2b_vision_decode_fp16_pf16+tower_fp16w32+chunk16: Pillow resize -> same tower .aimodelc -> same decoder .aimodelc, chunk order), text runs vs the round-6 b1 transcript (identical inputs). |dp| over the question's letters; bit-equality of the fp16 logits and of the tower's float32 output","all":{"slots":108,"max_abs_dp":6.298730608367276e-05,"mean_abs_dp_of_slot_max":1.0445199354181786e-06,"letter_logits_bit_equal":104,"full_logits_bit_equal":104,"full_logits_max_abs_diff":0.0078125,"argmax_equal":108,"tower_embeds_bit_equal_runs":68,"tower_embeds_max_abs_diff":0.053189992904663086},"by_arm":{"g256":{"slots":49,"max_abs_dp":0.0,"mean_abs_dp_of_slot_max":0.0,"letter_logits_bit_equal":49,"full_logits_bit_equal":49,"full_logits_max_abs_diff":0.0,"argmax_equal":49,"tower_embeds_bit_equal_runs":35,"tower_embeds_max_abs_diff":0.0},"g448":{"slots":49,"max_abs_dp":6.298730608367276e-05,"mean_abs_dp_of_slot_max":2.302207204595169e-06,"letter_logits_bit_equal":45,"full_logits_bit_equal":45,"full_logits_max_abs_diff":0.0078125,"argmax_equal":49,"tower_embeds_bit_equal_runs":33,"tower_embeds_max_abs_diff":0.053189992904663086},"text":{"slots":10,"max_abs_dp":0.0,"mean_abs_dp_of_slot_max":0.0,"letter_logits_bit_equal":10,"full_logits_bit_equal":10,"full_logits_max_abs_diff":0.0,"argmax_equal":10,"tower_embeds_bit_equal_runs":0,"tower_embeds_max_abs_diff":0.0}}},"python_bundle_records":{"e2e":"418213bfe687110f0169c95256951b0c7816f9d32521492c29c5a9503c396e25","b1":"418213bfe687110f0169c95256951b0c7816f9d32521492c29c5a9503c396e25"},"runs":[{"id":"r01","arm":"g256","tokens":105,"max_abs_dp":3.1101176778403072e-06,"mean_abs_dp":2.0505070706339357e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r01","arm":"g448","tokens":237,"max_abs_dp":5.660508126337227e-06,"mean_abs_dp":3.782820388258727e-06,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g256","tokens":105,"max_abs_dp":5.76876223912226e-06,"mean_abs_dp":3.861703080346353e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g448","tokens":237,"max_abs_dp":6.7384874481435375e-06,"mean_abs_dp":4.491306331423429e-06,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g256","tokens":105,"max_abs_dp":3.2721829886783738e-06,"mean_abs_dp":2.1677765254750196e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g448","tokens":237,"max_abs_dp":1.952898516088819e-05,"mean_abs_dp":1.3041442351732739e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g256","tokens":153,"max_abs_dp":3.77586493442994e-05,"mean_abs_dp":1.0321872803515752e-05,"calls":{"main":9,"prefill":9,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g448","tokens":285,"max_abs_dp":1.80627241554987e-05,"mean_abs_dp":5.892581068105695e-06,"calls":{"main":13,"prefill":17,"total":30},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g256","tokens":105,"max_abs_dp":5.2574588721099325e-05,"mean_abs_dp":3.502931432764067e-05,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g448","tokens":237,"max_abs_dp":3.062555064470285e-05,"mean_abs_dp":2.0415559169017123e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g256","tokens":126,"max_abs_dp":3.7307297401367734e-05,"mean_abs_dp":9.335266280162274e-06,"calls":{"main":14,"prefill":7,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g448","tokens":258,"max_abs_dp":8.912485169099149e-06,"mean_abs_dp":2.2223350869954436e-06,"calls":{"main":2,"prefill":16,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g256","tokens":184,"max_abs_dp":0.00011322475813147631,"mean_abs_dp":3.3781003194402385e-05,"calls":{"main":24,"prefill":10,"total":34},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g448","tokens":316,"max_abs_dp":0.00023937502200033212,"mean_abs_dp":5.175489972942714e-05,"calls":{"main":28,"prefill":18,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g256","tokens":120,"max_abs_dp":4.2761661226076164e-05,"mean_abs_dp":1.4249846493557126e-05,"calls":{"main":8,"prefill":7,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g448","tokens":252,"max_abs_dp":2.6698863657581562e-05,"mean_abs_dp":8.907014804793433e-06,"calls":{"main":12,"prefill":15,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g256","tokens":128,"max_abs_dp":0.0007433753117671582,"mean_abs_dp":0.00018583890211848738,"calls":{"main":0,"prefill":8,"total":8},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g448","tokens":260,"max_abs_dp":0.0009327309371999215,"mean_abs_dp":0.00023318194122061122,"calls":{"main":4,"prefill":16,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g256","tokens":112,"max_abs_dp":1.6713512907773342e-05,"mean_abs_dp":1.1182165679911204e-05,"calls":{"main":0,"prefill":7,"total":7},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g448","tokens":244,"max_abs_dp":6.150812051575194e-06,"mean_abs_dp":4.0899597000362e-06,"calls":{"main":4,"prefill":15,"total":19},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g256","tokens":103,"max_abs_dp":2.349873020557336e-05,"mean_abs_dp":2.3470208451782573e-05,"calls":{"main":7,"prefill":6,"total":13},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g448","tokens":235,"max_abs_dp":0.00012558110900007247,"mean_abs_dp":0.00012555689461311295,"calls":{"main":11,"prefill":14,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g256","tokens":117,"max_abs_dp":1.0879741226629434e-06,"mean_abs_dp":7.070113600252197e-07,"calls":{"main":5,"prefill":7,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g448","tokens":249,"max_abs_dp":2.131206145228326e-06,"mean_abs_dp":1.1544797896835405e-06,"calls":{"main":9,"prefill":15,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g256","tokens":140,"max_abs_dp":0.00038593584982181406,"mean_abs_dp":8.576149097778633e-05,"calls":{"main":12,"prefill":8,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g448","tokens":272,"max_abs_dp":4.504693712381602e-06,"mean_abs_dp":1.000640039981361e-06,"calls":{"main":0,"prefill":17,"total":17},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g256","tokens":111,"max_abs_dp":5.750033446938474e-06,"mean_abs_dp":3.146645146748537e-06,"calls":{"main":15,"prefill":6,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g448","tokens":243,"max_abs_dp":1.2019331816048862e-05,"mean_abs_dp":6.000398156959426e-06,"calls":{"main":3,"prefill":15,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g256","tokens":147,"max_abs_dp":4.112909655773489e-05,"mean_abs_dp":1.7471410993673965e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g448","tokens":279,"max_abs_dp":0.00012136859331268735,"mean_abs_dp":6.24758188288078e-05,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g256","tokens":193,"max_abs_dp":9.636731990181335e-05,"mean_abs_dp":2.215416722045421e-05,"calls":{"main":17,"prefill":11,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g448","tokens":325,"max_abs_dp":0.00012604835580953822,"mean_abs_dp":2.82599430498649e-05,"calls":{"main":21,"prefill":19,"total":40},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g256","tokens":171,"max_abs_dp":2.073876051555826e-05,"mean_abs_dp":8.168700370686157e-06,"calls":{"main":27,"prefill":9,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g448","tokens":303,"max_abs_dp":2.6149319381770647e-05,"mean_abs_dp":8.809094553737935e-06,"calls":{"main":15,"prefill":18,"total":33},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g256","tokens":211,"max_abs_dp":0.0008890173952338376,"mean_abs_dp":0.00010920418066449597,"calls":{"main":35,"prefill":11,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g448","tokens":343,"max_abs_dp":0.0030785221919065187,"mean_abs_dp":0.0003725329200313297,"calls":{"main":23,"prefill":20,"total":43},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g256","tokens":177,"max_abs_dp":0.00033381953355049454,"mean_abs_dp":8.493597399020465e-05,"calls":{"main":33,"prefill":9,"total":42},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g448","tokens":309,"max_abs_dp":4.186292624119847e-05,"mean_abs_dp":1.4741149222329405e-05,"calls":{"main":21,"prefill":18,"total":39},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.053189992904663086},{"id":"r20","arm":"g256","tokens":143,"max_abs_dp":0.002927743999633692,"mean_abs_dp":0.0005855483313550629,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r20","arm":"g448","tokens":275,"max_abs_dp":0.0021595096189871432,"mean_abs_dp":0.0004319065163818761,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g256","tokens":143,"max_abs_dp":8.336698959632294e-05,"mean_abs_dp":1.7045428912568975e-05,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g448","tokens":275,"max_abs_dp":3.370909396349653e-05,"mean_abs_dp":6.740573148775609e-06,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g256","tokens":147,"max_abs_dp":5.434280119354096e-05,"mean_abs_dp":1.2088469175174433e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g448","tokens":279,"max_abs_dp":0.0003145324254553161,"mean_abs_dp":6.989040959739295e-05,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.04187512397766113},{"id":"r23","arm":"g256","tokens":150,"max_abs_dp":6.326714395794224e-07,"mean_abs_dp":4.1541389043079044e-07,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r23","arm":"g448","tokens":282,"max_abs_dp":3.6570245875999774e-07,"mean_abs_dp":2.2558082231054327e-07,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g256","tokens":150,"max_abs_dp":8.540901203650303e-08,"mean_abs_dp":6.106723463345861e-08,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g448","tokens":282,"max_abs_dp":4.4172430135880347e-07,"mean_abs_dp":2.8186271906293373e-07,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g256","tokens":150,"max_abs_dp":6.44044415542e-05,"mean_abs_dp":4.2931560146337984e-05,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g448","tokens":282,"max_abs_dp":7.025503080948514e-05,"mean_abs_dp":4.684161545497519e-05,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g256","tokens":145,"max_abs_dp":0.00030462107334494615,"mean_abs_dp":0.0001523165320615591,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g448","tokens":277,"max_abs_dp":0.00854564497937682,"mean_abs_dp":0.004272811779478853,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g256","tokens":145,"max_abs_dp":0.0005691155425143757,"mean_abs_dp":0.00028455462804349765,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g448","tokens":277,"max_abs_dp":0.0009406553394049055,"mean_abs_dp":0.0004703336068838807,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g256","tokens":170,"max_abs_dp":0.0014256226063197042,"mean_abs_dp":0.00047520244226959343,"calls":{"main":10,"prefill":10,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g448","tokens":302,"max_abs_dp":0.0065295452356902706,"mean_abs_dp":0.0021765126435429227,"calls":{"main":14,"prefill":18,"total":32},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g256","tokens":164,"max_abs_dp":0.0012149093184980475,"mean_abs_dp":0.0007253338612302337,"calls":{"main":4,"prefill":10,"total":14},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g448","tokens":296,"max_abs_dp":0.0003930164465574121,"mean_abs_dp":0.00019649896826064845,"calls":{"main":8,"prefill":18,"total":26},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g256","tokens":121,"max_abs_dp":3.362448612986846e-06,"mean_abs_dp":1.6616574374654584e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g448","tokens":253,"max_abs_dp":2.437358188056571e-06,"mean_abs_dp":1.2044327689858431e-06,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g256","tokens":114,"max_abs_dp":0.00011723796896412342,"mean_abs_dp":5.86270535190787e-05,"calls":{"main":2,"prefill":7,"total":9},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g448","tokens":246,"max_abs_dp":0.0002686848366134642,"mean_abs_dp":0.00013434199630119398,"calls":{"main":6,"prefill":15,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g256","tokens":105,"max_abs_dp":7.360306083403767e-05,"mean_abs_dp":7.358420155190084e-05,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g448","tokens":237,"max_abs_dp":7.291911551599028e-05,"mean_abs_dp":7.291399324182903e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g256","tokens":144,"max_abs_dp":3.32742540922748e-05,"mean_abs_dp":1.87112203890406e-05,"calls":{"main":16,"prefill":8,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g448","tokens":276,"max_abs_dp":1.4841931224585103e-05,"mean_abs_dp":7.486076748235324e-06,"calls":{"main":20,"prefill":16,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g256","tokens":121,"max_abs_dp":1.811342492796797e-05,"mean_abs_dp":5.178441779765734e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g448","tokens":253,"max_abs_dp":6.350772148278061e-06,"mean_abs_dp":1.8125709233494673e-06,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g256","tokens":145,"max_abs_dp":0.00016211157654966168,"mean_abs_dp":5.631958124214251e-05,"calls":{"main":17,"prefill":8,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g448","tokens":277,"max_abs_dp":0.00014301157085283567,"mean_abs_dp":4.8063393245812314e-05,"calls":{"main":21,"prefill":16,"total":37},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"t01","arm":"text","tokens":62,"max_abs_dp":1.1261840832110792e-06,"mean_abs_dp":7.590451753452418e-07,"calls":{"main":14,"prefill":3,"total":17},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t02","arm":"text","tokens":107,"max_abs_dp":9.973106485561267e-05,"mean_abs_dp":4.0770005194156294e-05,"calls":{"main":27,"prefill":5,"total":32},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t03","arm":"text","tokens":51,"max_abs_dp":0.00022168741880501663,"mean_abs_dp":8.867790118815034e-05,"calls":{"main":3,"prefill":3,"total":6},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t04","arm":"text","tokens":74,"max_abs_dp":0.0007736432970376628,"mean_abs_dp":0.00036234723508736207,"calls":{"main":26,"prefill":3,"total":29},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t05","arm":"text","tokens":78,"max_abs_dp":3.89124183164391e-06,"mean_abs_dp":1.1100860034822475e-06,"calls":{"main":14,"prefill":4,"total":18},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t06","arm":"text","tokens":95,"max_abs_dp":2.4181264343321393e-05,"mean_abs_dp":1.1471266469224055e-05,"calls":{"main":15,"prefill":5,"total":20},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null}],"slots":[{"id":"r01","arm":"g256","t":104,"nopts":3,"probs":[0.9997497094204386,0.0001392615731653799,0.00011102900639593486],"probs_oracle":[0.9997528195381165,0.00013744122406933457,0.0001098079519579187],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.1101176778403072e-06,"mean_abs_dp":2.0505070706339357e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996153783140471,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r01","arm":"g448","t":236,"nopts":3,"probs":[0.9996612687368688,0.00018519730482376682,0.00015353395830743424],"probs_oracle":[0.9996669292449951,0.00018169928807765245,0.00015134402201510966],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.660508126337227e-06,"mean_abs_dp":3.782820388258727e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9994852299569175,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r02","arm":"g256","t":104,"nopts":3,"probs":[0.00019868280241103374,0.9996366033691085,0.00016471382848043005],"probs_oracle":[0.00019598512153606862,0.9996423721313477,0.00016159516235347837],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.76876223912226e-06,"mean_abs_dp":3.861703080346353e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9994463870098116,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r02","arm":"g448","t":236,"nopts":3,"probs":[0.000273651307111891,0.9994692770803314,0.0002570716125568927],"probs_oracle":[0.000269706710241735,0.9994760155677795,0.00025428077788092196],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.7384874481435375e-06,"mean_abs_dp":4.491306331423429e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992063088575378,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r03","arm":"g256","t":104,"nopts":3,"probs":[0.00026939816754023544,0.0003005353733280939,0.9994300664591317],"probs_oracle":[0.00026771891862154007,0.0002989834756590426,0.9994333386421204],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.2721829886783738e-06,"mean_abs_dp":2.1677765254750196e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991343551664613,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r03","arm":"g448","t":236,"nopts":3,"probs":[0.0004440050968048193,0.0004800839774673775,0.9990759109257278],"probs_oracle":[0.00043719500536099076,0.000467298727016896,0.9990954399108887],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.952898516088819e-05,"mean_abs_dp":1.3041442351732739e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9986281411838718,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g256","t":112,"nopts":4,"probs":[0.9992263244891981,0.00044755484238076954,0.00024333134714230173,8.278932127888766e-05],"probs_oracle":[0.9992350339889526,0.0004433949070516974,0.0002398906071903184,8.168041676981375e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.709499754511185e-06,"mean_abs_dp":4.3547698861601464e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9987916390819009,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g256","t":152,"nopts":5,"probs":[0.9963341066231777,0.001418209484759651,0.0009228465125380186,0.0008534935019014999,0.00047134387762331987],"probs_oracle":[0.996371865272522,0.0014012253377586603,0.0009149523684754968,0.0008468246669508517,0.0004651718772947788],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.77586493442994e-05,"mean_abs_dp":1.5095555137400237e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9949706399347633,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g448","t":244,"nopts":4,"probs":[0.9994266676448472,0.0003149572959401018,0.0001739359383796407,8.443912083312718e-05],"probs_oracle":[0.9994351267814636,0.00030979173607192934,0.0001716899569146335,8.336015162058175e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.459136616401075e-06,"mean_abs_dp":4.237411790531538e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991253350453917,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r04","arm":"g448","t":284,"nopts":5,"probs":[0.9983501409158286,0.0006256926877117296,0.00048349783463579147,0.000337535886950096,0.00020313267487369043],"probs_oracle":[0.9983682036399841,0.0006189235136844218,0.0004789685772266239,0.0003326288133393973,0.00020131732162553817],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.80627241554987e-05,"mean_abs_dp":7.216716490165021e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9977492801262997,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r05","arm":"g256","t":104,"nopts":3,"probs":[0.00567870238070768,0.9930542078476498,0.0012670897716425744],"probs_oracle":[0.005630022846162319,0.9931067824363708,0.0012632559519261122],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.2574588721099325e-05,"mean_abs_dp":3.502931432764067e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9874767595902085,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r05","arm":"g448","t":236,"nopts":3,"probs":[0.005630313487890581,0.9923145287347862,0.002055157777323089],"probs_oracle":[0.005602917168289423,0.9923451542854309,0.0020519329700618982],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.062555064470285e-05,"mean_abs_dp":2.0415559169017123e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9867422371171415,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r06","arm":"g256","t":125,"nopts":8,"probs":[0.0004539557896578944,0.00043316764054686954,0.9978041700967026,0.00014737786718404345,0.00019524395221938779,0.00024109409999459213,0.0006204839969199214,0.00010450655677464298],"probs_oracle":[0.00044743184116669,0.00042624998604878783,0.997841477394104,0.0001445232192054391,0.00019250228069722652,0.00023652952222619206,0.0006083882763050497,0.00010282994480803609],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.7307297401367734e-05,"mean_abs_dp":9.335266280162274e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.997233089117799,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r06","arm":"g448","t":257,"nopts":8,"probs":[0.0004832139490430726,0.00044689984730200135,0.9977639599627984,0.00015810699924844444,0.00016569471256822025,0.00023920825067178786,0.000606086075193931,0.00013683020317421904],"probs_oracle":[0.00048085491289384663,0.0004459824995137751,0.9977728724479675,0.0001569857558934018,0.00016541295917704701,0.00023827303084544837,0.0006033360259607434,0.00013632865739054978],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.912485169099149e-06,"mean_abs_dp":2.2223350869954436e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9971695364220068,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":112,"nopts":4,"probs":[0.9987509535327098,0.0005880198646746591,0.0002568868539085187,0.0004041397487070115],"probs_oracle":[0.9987712502479553,0.00058116385480389,0.0002529850753489882,0.000394599192077294],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.0296715245504693e-05,"mean_abs_dp":1.0148765076380462e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9981900863931514,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":156,"nopts":6,"probs":[0.0016061101786271905,0.9957564679004233,0.0005464514668958667,0.0005507373388693371,0.000989495776314906,0.0005507373388693371],"probs_oracle":[0.0015793738421052694,0.9958256483078003,0.0005403549876064062,0.0005451973993331194,0.0009696456836536527,0.0005397982313297689],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.918040737702302e-05,"mean_abs_dp":2.305706048757398e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.994246274465695,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g256","t":183,"nopts":2,"probs":[0.060529390202538995,0.9394706097974611],"probs_oracle":[0.06064261496067047,0.9393573999404907],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00011322475813147631,"mean_abs_dp":0.00011321730755093143,"finite":true,"read_from":"main","oracle_top2_margin":0.8787147849798203,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":244,"nopts":4,"probs":[0.9985431764562523,0.000597155573887962,0.00030738934604395167,0.0005522786238157235],"probs_oracle":[0.9985532164573669,0.0005945708835497499,0.0003056767163798213,0.0005464947316795588],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0040001114597352e-05,"mean_abs_dp":5.0303033132761586e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9979586455738172,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":288,"nopts":6,"probs":[0.0016958621127589775,0.9954463530755453,0.0006336973734950824,0.0006094208083522638,0.001028592366033363,0.000586074263815153],"probs_oracle":[0.0016743798041716218,0.9955074787139893,0.0006255566840991378,0.0006036655395291746,0.001013218774460256,0.0005756971659138799],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.112563844395957e-05,"mean_abs_dp":2.0375765787454855e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9938330989098176,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r07","arm":"g448","t":315,"nopts":2,"probs":[0.07211220541293051,0.9278877945870695],"probs_oracle":[0.07187289744615555,0.9281271696090698],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00023937502200033212,"mean_abs_dp":0.00023934149438764596,"finite":true,"read_from":"main","oracle_top2_margin":0.8562542721629143,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r08","arm":"g256","t":119,"nopts":6,"probs":[0.0006815973774667356,0.0005434168081121444,0.9979930864794905,0.00039140844373680765,0.00013579655924772193,0.0002546943319460747],"probs_oracle":[0.0006678571808151901,0.0005291247507557273,0.9980358481407166,0.0003851840447168797,0.00013258290709927678,0.00024942721938714385],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.2761661226076164e-05,"mean_abs_dp":1.4249846493557126e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9973679909599014,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r08","arm":"g448","t":251,"nopts":6,"probs":[0.0005697310728025495,0.0004873168109435057,0.9984051625369467,0.00026288770068428734,9.978091647964578e-05,0.00017512096214336893],"probs_oracle":[0.0005603667814284563,0.00047818213352002203,0.9984318614006042,0.00025924580404534936,9.779584797797725e-05,0.00017250367091037333],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.6698863657581562e-05,"mean_abs_dp":8.907014804793433e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9978714946191758,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r09","arm":"g256","t":127,"nopts":8,"probs":[0.0007091801456934003,0.0006981853266301289,0.000569842255876744,0.0177272729702159,0.975467457284058,0.003715838277693117,0.0007849908631788027,0.00032723287665389135],"probs_oracle":[0.0006885188049636781,0.0006795556400902569,0.0005544525920413435,0.01720018871128559,0.9762108325958252,0.0035912375897169113,0.0007585337734781206,0.0003167196991853416],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0007433753117671582,"mean_abs_dp":0.00018583890211848738,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9590106438845396,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r09","arm":"g448","t":259,"nopts":8,"probs":[0.001168663579751714,0.000939047567378768,0.0006816754299582218,0.002200475242664631,0.9524007065055796,0.039310271908001494,0.0025327314792745947,0.0007664282873910434],"probs_oracle":[0.0011424915865063667,0.0009199993801303208,0.0006675715558230877,0.0021507099736481905,0.9533334374427795,0.03857970982789993,0.0024612408597022295,0.0007448457181453705],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0009327309371999215,"mean_abs_dp":0.00023318194122061122,"finite":true,"read_from":"main","oracle_top2_margin":0.9147537276148796,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r10","arm":"g256","t":111,"nopts":3,"probs":[0.9985505100359814,0.00114209836042169,0.0003073916035970853],"probs_oracle":[0.9985672235488892,0.0011292636627331376,0.00030339331715367734],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.6713512907773342e-05,"mean_abs_dp":1.1182165679911204e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.997437959886156,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r10","arm":"g448","t":243,"nopts":3,"probs":[0.9995549801884672,0.00033270391127001025,0.00011231590026279501],"probs_oracle":[0.9995611310005188,0.0003284418198745698,0.00011045892460970208],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.150812051575194e-06,"mean_abs_dp":4.0899597000362e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992326891806442,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r11","arm":"g256","t":102,"nopts":2,"probs":[0.002800926967120973,0.9971990730328789],"probs_oracle":[0.002824368653818965,0.9971755743026733],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.349873020557336e-05,"mean_abs_dp":2.3470208451782573e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9943512056488544,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r11","arm":"g448","t":234,"nopts":2,"probs":[0.007176881945656949,0.9928231180543431],"probs_oracle":[0.007302414625883102,0.992697536945343],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012558110900007247,"mean_abs_dp":0.00012555689461311295,"finite":true,"read_from":"main","oracle_top2_margin":0.9853951223194599,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r12","arm":"g256","t":116,"nopts":5,"probs":[0.9988258564791708,0.0006663637713119991,0.00030508356558258374,0.00011994004280138088,8.275614113347764e-05],"probs_oracle":[0.9988265633583069,0.0006655653123743832,0.0003061715397052467,0.00011903217091457918,8.272226841654629e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0879741226629434e-06,"mean_abs_dp":7.070113600252197e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9981609980459325,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r12","arm":"g448","t":248,"nopts":5,"probs":[0.9991537478145457,0.0004581350215318546,0.00020649763251259106,0.00010505748720281517,7.656204420689791e-05],"probs_oracle":[0.9991558790206909,0.0004565727722365409,0.00020724281785078347,0.00010403238411527127,7.625338912475854e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.131206145228326e-06,"mean_abs_dp":1.1544797896835405e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986993062484544,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r13","arm":"g256","t":139,"nopts":9,"probs":[6.69528218381079e-05,0.00012314508581246774,9.665768533706018e-05,4.0927439056679186e-05,0.00012411092469099131,0.00017778069136575113,0.00021444444768834437,0.016771322616064302,0.9823846582881463],"probs_oracle":[6.637163460254669e-05,0.00012154777505202219,9.527682414045557e-05,4.0484348573954776e-05,0.00012243450328242034,0.00017381930956616998,0.00021287811978254467,0.017157258465886116,0.9820099472999573],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00038593584982181406,"mean_abs_dp":8.576149097778633e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9648526888340712,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r13","arm":"g448","t":271,"nopts":9,"probs":[2.045729378346038e-05,3.320546540900632e-05,3.399291032864415e-05,9.220824724934043e-06,2.75282977067983e-05,3.346589910199451e-05,4.50332895082248e-05,9.312703074569437e-05,0.9997039689886912],"probs_oracle":[2.0277637304388918e-05,3.2944568374659866e-05,3.3252264984184876e-05,9.089779268833809e-06,2.7153426344739273e-05,3.251702946727164e-05,4.40161456936039e-05,9.227909322362393e-05,0.9997084736824036],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.504693712381602e-06,"mean_abs_dp":1.000640039981361e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9996161945891799,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r14","arm":"g256","t":110,"nopts":4,"probs":[0.9989199918242184,0.0006459227833106589,0.0001850599758197889,0.00024902541665109345],"probs_oracle":[0.9989142417907715,0.0006516185821965337,0.0001856259477790445,0.00024845064035616815],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.750033446938474e-06,"mean_abs_dp":3.146645146748537e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.998262623208575,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r14","arm":"g448","t":242,"nopts":4,"probs":[0.9992450087710587,0.0004653921025630391,0.00012235697313286097,0.0001672421532451772],"probs_oracle":[0.9992570281028748,0.0004573110491037369,0.00012069327203789726,0.00016500464698765427],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2019331816048862e-05,"mean_abs_dp":6.000398156959426e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.998799717053771,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g256","t":114,"nopts":4,"probs":[0.0013225919568139097,0.000960098798122708,0.996843129341424,0.0008741799036392433],"probs_oracle":[0.0013125522527843714,0.0009568282985128462,0.9968631267547607,0.00086748949252069],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.9997413336692027e-05,"mean_abs_dp":9.999507023661348e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9955505745019764,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g256","t":146,"nopts":3,"probs":[0.9973401408352355,0.001940419867309761,0.0007194392974546785],"probs_oracle":[0.9973812699317932,0.001912115840241313,0.000706570572219789],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.112909655773489e-05,"mean_abs_dp":2.7433949620357456e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9954691540915519,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g448","t":246,"nopts":4,"probs":[0.006672715868641399,0.003329141469421999,0.982612119489573,0.007386023172363682],"probs_oracle":[0.006657375022768974,0.0033173311967402697,0.9827334880828857,0.007291808724403381],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012136859331268735,"mean_abs_dp":6.0683539956785516e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9754416793584824,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r15","arm":"g448","t":278,"nopts":3,"probs":[0.9946688913143538,0.004395296136690761,0.0009358125489554722],"probs_oracle":[0.9947662353515625,0.004311799071729183,0.0009220570791512728],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.73440372087353e-05,"mean_abs_dp":6.48655239915042e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9904544362798333,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":117,"nopts":4,"probs":[0.0007782375337297331,0.0010889500900131026,0.9977737147500348,0.00035909762622219493],"probs_oracle":[0.0007772505050525069,0.0010821042815223336,0.997784435749054,0.0003561374032869935],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0720999019109279e-05,"mean_abs_dp":5.378514780576485e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9967023314675316,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":165,"nopts":7,"probs":[0.0015806405637158872,0.0038514711221036527,0.9876517787372271,0.0038514711221036527,0.0010611918730116016,0.0010447396179590175,0.0009587069638791459],"probs_oracle":[0.001566192484460771,0.0037890495732426643,0.9877481460571289,0.003863656660541892,0.0010514376917853951,0.0010348198702558875,0.0009466473711654544],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.636731990181335e-05,"mean_abs_dp":3.102228687131217e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.983884489396587,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g256","t":192,"nopts":2,"probs":[0.003621685478675127,0.996378314521325],"probs_oracle":[0.0035970157478004694,0.9964029788970947],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.4669730874657403e-05,"mean_abs_dp":2.466705332220681e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9928059631492943,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":249,"nopts":4,"probs":[0.00016921486556775047,0.0004304206779588754,0.9992527726275144,0.0001475918289590957],"probs_oracle":[0.00016730134666431695,0.00042254829895682633,0.9992651343345642,0.00014504071441479027],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2361707049768356e-05,"mean_abs_dp":6.174679874889092e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9988425860356074,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":297,"nopts":7,"probs":[0.001246679207623153,0.0026186813310170136,0.9924442922775828,0.0017308426698170456,0.0006028546731408958,0.0007048085187285085,0.0006518413220907323],"probs_oracle":[0.0012208223342895508,0.0025618993677198887,0.9925703406333923,0.0017182007431983948,0.0005935655790381134,0.0006946290959604084,0.0006406526081264019],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012604835580953822,"mean_abs_dp":3.5998049984875576e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9900084412656724,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r16","arm":"g448","t":324,"nopts":2,"probs":[0.0044683842664911206,0.9955316157335089],"probs_oracle":[0.004423052072525024,0.9955769777297974],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.536199628846216e-05,"mean_abs_dp":4.534709512727915e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9911539256572723,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":108,"nopts":3,"probs":[0.9997872736204506,0.00014538135275247673,6.73450267968826e-05],"probs_oracle":[0.9997918009757996,0.00014244449266698211,6.58148928778246e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.5273553489488805e-06,"mean_abs_dp":2.9981164511671643e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996493564831326,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":141,"nopts":3,"probs":[0.9992306625208732,0.0004011873095648438,0.0003681501695620029],"probs_oracle":[0.999238133430481,0.00039801813545636833,0.0003638898197095841],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.47090960773189e-06,"mean_abs_dp":4.966811189542062e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9988401152950246,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g256","t":170,"nopts":2,"probs":[0.9981324340221067,0.0018675659778932805],"probs_oracle":[0.9981531500816345,0.0018468272173777223],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.073876051555826e-05,"mean_abs_dp":2.0727410021680786e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9963063228642568,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":240,"nopts":3,"probs":[0.9998059829765886,0.00013341192497696007,6.0605098434396695e-05],"probs_oracle":[0.9998090863227844,0.0001311514643020928,5.978461922495626e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.103346195798551e-06,"mean_abs_dp":2.0614286933687555e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996779348584823,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":273,"nopts":3,"probs":[0.9991684533091557,0.00043038435804146596,0.0004011623328030057],"probs_oracle":[0.9991744160652161,0.00042735174065455794,0.0003981590270996094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.9627560603514596e-06,"mean_abs_dp":3.999559716885263e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9987470643245615,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r17","arm":"g448","t":302,"nopts":2,"probs":[0.997965182999252,0.002034817000747976],"probs_oracle":[0.9979913234710693,0.002008667681366205],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.6149319381770647e-05,"mean_abs_dp":2.6144895599570714e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9959826557897031,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":123,"nopts":6,"probs":[0.9988501756008867,0.0006768739537244369,0.0001711405368704473,7.360662770424236e-05,0.00012520909152073824,0.00010299418929328017],"probs_oracle":[0.9988755583763123,0.000659995072055608,0.00016870087711140513,7.21013275324367e-05,0.00012259691720828414,0.00010095535981236026],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.538277542551981e-05,"mean_abs_dp":8.47627013642842e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9982155633042566,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":164,"nopts":5,"probs":[0.0032340509923870107,0.015918556888829562,0.9471173741300347,0.03044503810377914,0.003284979884969603],"probs_oracle":[0.0031590552534908056,0.015553153119981289,0.9480063915252686,0.03006896562874317,0.0032123548444360495],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0008890173952338376,"mean_abs_dp":0.0003556228837095674,"finite":true,"read_from":"main","oracle_top2_margin":0.9179374258965254,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g256","t":210,"nopts":6,"probs":[0.0005261191787456162,0.000909052422139744,0.0005428199942296315,0.0004753090683425212,0.9968970288681479,0.0006496704683945269],"probs_oracle":[0.0005212401156313717,0.0009053464164026082,0.000543770962394774,0.0004741615557577461,0.9969097971916199,0.0006456233095377684],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2768323471967769e-05,"mean_abs_dp":4.5831719883373606e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9960044507752173,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":255,"nopts":6,"probs":[0.9989476949483892,0.0004509392811537855,0.00020167252103176967,8.844902955360989e-05,0.0001504568675635731,0.00016078735230808905],"probs_oracle":[0.9989650249481201,0.00044237051042728126,0.00019978830823674798,8.695968426764011e-05,0.0001475344179198146,0.0001582982367835939],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.7329999730919532e-05,"mean_abs_dp":5.780648951111485e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9985226544376928,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":296,"nopts":5,"probs":[0.008354139503804588,0.046595648874413455,0.9213888626233095,0.01882634398905573,0.004835005009416626],"probs_oracle":[0.008088547736406326,0.04423723742365837,0.9244673848152161,0.018492545932531357,0.004714324604719877],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0030785221919065187,"mean_abs_dp":0.001231400774256197,"finite":true,"read_from":"main","oracle_top2_margin":0.8802301473915577,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r18","arm":"g448","t":342,"nopts":6,"probs":[0.0006442081989388584,0.0008736758834440874,0.0005774644973956571,0.0005056448089159738,0.9962683860117058,0.00113062059959965],"probs_oracle":[0.0006307050934992731,0.0008567753247916698,0.0005710572004318237,0.0004954338655807078,0.9963390827178955,0.0011069673346355557],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.069670618975454e-05,"mean_abs_dp":2.356197925749184e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.99523211538326,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":110,"nopts":3,"probs":[0.9929774240680236,0.0059044595935782325,0.0011181163383981352],"probs_oracle":[0.9930190443992615,0.005865252576768398,0.0011157557601109147],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.162033123789044e-05,"mean_abs_dp":2.7729308778315046e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9871537918224931,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":146,"nopts":4,"probs":[0.0024414779146855314,0.9849624894607427,0.011738993271029067,0.0008570393535427288],"probs_oracle":[0.0024147704243659973,0.9852963089942932,0.011448029428720474,0.0008408919093199074],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00033381953355049454,"mean_abs_dp":0.00016690957760036066,"finite":true,"read_from":"main","oracle_top2_margin":0.9738482795655727,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g256","t":176,"nopts":2,"probs":[0.997965182999252,0.002034817000747976],"probs_oracle":[0.9979719519615173,0.0020279884338378906],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.8285669100852375e-06,"mean_abs_dp":6.7987645877270325e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9959439635276794,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r19","arm":"g448","t":242,"nopts":3,"probs":[0.9957034887714526,0.0036476185814633337,0.0006488926470837863],"probs_oracle":[0.9957261085510254,0.0036251493729650974,0.0006488442304544151],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.2619779572741194e-05,"mean_abs_dp":1.5045801566782881e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9921009591780603,"python":{"ref":"e2e","max_abs_dp":3.355204093569153e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":278,"nopts":4,"probs":[0.0007254553211672244,0.9978537354853189,0.001196073618950996,0.00022473557456311583],"probs_oracle":[0.000710506399627775,0.9978955984115601,0.0011723366333171725,0.00022145592083688825],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.186292624119847e-05,"mean_abs_dp":2.0957121785174714e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9967232617782429,"python":{"ref":"e2e","max_abs_dp":1.7622237268264947e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":308,"nopts":2,"probs":[0.9981469406043857,0.0018530593956144392],"probs_oracle":[0.9981451034545898,0.0018549266969785094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.867301364070215e-06,"mean_abs_dp":1.8522255799585706e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9962901767576113,"python":{"ref":"e2e","max_abs_dp":1.4506582278972502e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r20","arm":"g256","t":142,"nopts":10,"probs":[0.00038087895168696287,0.0010272792819467234,0.0004418273136058572,0.0005899168604847742,0.0006843155300768766,0.00034951417281430593,0.0014946825356476332,0.46596071759017954,0.001068201436337015,0.5280026663272203],"probs_oracle":[0.0003781993000302464,0.0010252433130517602,0.00043770880438387394,0.0005875966744497418,0.0006762796547263861,0.00034810963552445173,0.0014829656574875116,0.46888846158981323,0.001062334282323718,0.5251131057739258],"probs_swift_json_equal":true,"argmax":9,"argmax_oracle":9,"argmax_equal":true,"full_vocab_top1_id":41,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.002927743999633692,"mean_abs_dp":0.0005855483313550629,"finite":true,"read_from":"main","oracle_top2_margin":0.05622464418411255,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r20","arm":"g448","t":274,"nopts":10,"probs":[0.000454145605581403,0.0010979832465064854,0.0006404479566379976,0.0007033944739160236,0.0009842258211508678,0.0004299764173287985,0.0021666075099919706,0.8878364825646982,0.0012937404350163266,0.10439299596917198],"probs_oracle":[0.0004485415993258357,0.0010800594463944435,0.0006290642777457833,0.0006908925715833902,0.000972151174210012,0.00042305333772674203,0.0021294124890118837,0.8899959921836853,0.0012687055859714746,0.10236208140850067],"probs_swift_json_equal":true,"argmax":7,"argmax_oracle":7,"argmax_equal":true,"full_vocab_top1_id":39,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0021595096189871432,"mean_abs_dp":0.0004319065163818761,"finite":true,"read_from":"main","oracle_top2_margin":0.7876339107751846,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r21","arm":"g256","t":142,"nopts":10,"probs":[0.00038346652675958496,0.0006889645604304637,0.00039563906098078003,0.9701343377833529,0.0009417029013204846,0.0004962425570249398,0.0005282474522322782,0.0007391510762796228,0.0004379325195051688,0.02525431556211353],"probs_oracle":[0.00038343475898727775,0.000686735613271594,0.0003942575422115624,0.9702177047729492,0.0009427570621483028,0.0004960547666996717,0.0005290430854074657,0.0007387973600998521,0.0004333727993071079,0.02517782151699066],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.336698959632294e-05,"mean_abs_dp":1.7045428912568975e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9450398832559586,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r21","arm":"g448","t":274,"nopts":10,"probs":[0.0005460758646117826,0.0008863686186938355,0.000550358790700028,0.9644568887204283,0.0012306009573601823,0.0006849325815120165,0.0006849325815120165,0.0010122647357908017,0.0005950797393426171,0.02935249741004818],"probs_oracle":[0.0005481093539856374,0.0008933480712585151,0.0005559168639592826,0.9644231796264648,0.0012373479548841715,0.0006897316197864711,0.0006905003683641553,0.0010131854796782136,0.0005959380650892854,0.02935273014008999],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.370909396349653e-05,"mean_abs_dp":6.740573148775609e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9350704494863749,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r22","arm":"g256","t":146,"nopts":9,"probs":[0.0015211552740545752,0.0019995208860519036,0.0010054222423193327,0.9883457588604886,0.0017508389882165889,0.0021451727706363164,0.0015451099840254866,0.0008270377772684362,0.0008599832169387924],"probs_oracle":[0.0015125663485378027,0.0019855706486850977,0.0009997023735195398,0.9884001016616821,0.0017508834134787321,0.0021429003681987524,0.0015369588509202003,0.0008162282756529748,0.0008550662896595895],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.434280119354096e-05,"mean_abs_dp":1.2088469175174433e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9862572012934834,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r22","arm":"g448","t":278,"nopts":9,"probs":[0.002518919534805213,0.0038408936218582265,0.001626336685837417,0.9849393380617578,0.00153978462623574,0.0017448045672751099,0.0018573347937589281,0.0007049647718337574,0.001227623336637637],"probs_oracle":[0.0024558305740356445,0.0037613804452121258,0.0016030978877097368,0.9852538704872131,0.0015132761327549815,0.0017176602268591523,0.0018075669649988413,0.0006852896767668426,0.0012020787689834833],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003145324254553161,"mean_abs_dp":6.989040959739295e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.981492490042001,"python":{"ref":"e2e","max_abs_dp":6.298730608367276e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r23","arm":"g256","t":149,"nopts":3,"probs":[0.9997600586115958,1.8267332179113375e-05,0.00022167405622496576],"probs_oracle":[0.9997594952583313,1.821711521188263e-05,0.00022230672766454518],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.326714395794224e-07,"mean_abs_dp":4.1541389043079044e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9995371885306668,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r23","arm":"g448","t":281,"nopts":3,"probs":[0.9997156343143667,2.3181421300410533e-05,0.0002611842643328929],"probs_oracle":[0.999715268611908,2.3231894374475814e-05,0.00026144483126699924],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.6570245875999774e-07,"mean_abs_dp":2.2558082231054327e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.999453823780641,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r24","arm":"g256","t":149,"nopts":3,"probs":[3.941037375793803e-05,0.9991438837095291,0.0008167059167129814],"probs_oracle":[3.9324964745901525e-05,0.9991439580917358,0.000816729327198118],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.540901203650303e-08,"mean_abs_dp":6.106723463345861e-08,"finite":true,"read_from":"main","oracle_top2_margin":0.9983272287645377,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r24","arm":"g448","t":281,"nopts":3,"probs":[3.789512539755053e-05,0.9989999287781706,0.000962176096431801],"probs_oracle":[3.7778951082145795e-05,0.9990003705024719,0.0009618884068913758],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.4172430135880347e-07,"mean_abs_dp":2.8186271906293373e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9980384820955805,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r25","arm":"g256","t":149,"nopts":3,"probs":[0.0012049803729183377,0.0013337916871965301,0.997461227939885],"probs_oracle":[0.001173934550024569,0.0013004472712054849,0.9975256323814392],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.44044415542e-05,"mean_abs_dp":4.2931560146337984e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9962251851102337,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r25","arm":"g448","t":281,"nopts":3,"probs":[0.0025878121618512708,0.0012223959100349578,0.9961897919281139],"probs_oracle":[0.0025366705376654863,0.0012032677186653018,0.9962600469589233],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.025503080948514e-05,"mean_abs_dp":4.684161545497519e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9937233764212579,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r26","arm":"g256","t":144,"nopts":4,"probs":[0.987037163588146,0.0012302380556085092,0.010381446180729453,0.0013511521755160186],"probs_oracle":[0.986732542514801,0.001274453941732645,0.01060215849429369,0.0013908690307289362],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00030462107334494615,"mean_abs_dp":0.0001523165320615591,"finite":true,"read_from":"main","oracle_top2_margin":0.9761303840205073,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r26","arm":"g448","t":276,"nopts":4,"probs":[0.8417490799235297,0.028803069605559367,0.0826958537706501,0.04675199670026068],"probs_oracle":[0.8332034349441528,0.030570058152079582,0.08590248227119446,0.050323981791734695],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00854564497937682,"mean_abs_dp":0.004272811779478853,"finite":true,"read_from":"main","oracle_top2_margin":0.7473009526729584,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r27","arm":"g256","t":144,"nopts":4,"probs":[0.00758993871643747,0.008600527315403133,0.005135609425834902,0.9786739245423245],"probs_oracle":[0.007379894610494375,0.008370750583708286,0.005006327293813229,0.9792430400848389],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0005691155425143757,"mean_abs_dp":0.00028455462804349765,"finite":true,"read_from":"main","oracle_top2_margin":0.9708722895011306,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r27","arm":"g448","t":276,"nopts":4,"probs":[0.010668233815658327,0.013485911171321044,0.00799013793879034,0.9678557170742303],"probs_oracle":[0.010312587022781372,0.013162228278815746,0.007728788536041975,0.9687963724136353],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0009406553394049055,"mean_abs_dp":0.0004703336068838807,"finite":true,"read_from":"main","oracle_top2_margin":0.9556341441348195,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r28","arm":"g256","t":169,"nopts":6,"probs":[0.03043050791961641,0.7909685088699719,0.17511548230643817,0.0012560170655561205,0.0008173058924599845,0.0014121779459575056],"probs_oracle":[0.029139939695596695,0.7908696532249451,0.17654110491275787,0.0012475387193262577,0.0008099511615000665,0.0013918428448960185],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0014256226063197042,"mean_abs_dp":0.00047520244226959343,"finite":true,"read_from":"main","oracle_top2_margin":0.6143285483121872,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r28","arm":"g448","t":301,"nopts":6,"probs":[0.024532423949363037,0.7339443212270172,0.2382768347900004,0.00125036207465923,0.000755426279162797,0.0012406316797971817],"probs_oracle":[0.023014351725578308,0.7404738664627075,0.23333954811096191,0.0012221261858940125,0.000736995309125632,0.001213126815855503],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0065295452356902706,"mean_abs_dp":0.0021765126435429227,"finite":true,"read_from":"main","oracle_top2_margin":0.5071343183517456,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r29","arm":"g256","t":163,"nopts":4,"probs":[0.008080510104721255,0.8705650663040364,0.058781665926510726,0.06257275766473142],"probs_oracle":[0.007844753563404083,0.8715009689331055,0.05929643288254738,0.06135784834623337],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0012149093184980475,"mean_abs_dp":0.0007253338612302337,"finite":true,"read_from":"main","oracle_top2_margin":0.8101431205868721,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r29","arm":"g448","t":295,"nopts":4,"probs":[0.002621658991004417,0.9781688662998542,0.01644045096245729,0.0027690237466841813],"probs_oracle":[0.002556135645136237,0.9778785705566406,0.016833467409014702,0.002731863409280777],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003930164465574121,"mean_abs_dp":0.00019649896826064845,"finite":true,"read_from":"main","oracle_top2_margin":0.9610451031476259,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r30","arm":"g256","t":120,"nopts":4,"probs":[0.9998176448031265,6.867537456527629e-05,6.204303059341216e-05,5.1636791714740355e-05],"probs_oracle":[0.9998210072517395,6.738237425452098e-05,6.102487168391235e-05,5.0663769798120484e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.362448612986846e-06,"mean_abs_dp":1.6616574374654584e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999753624877485,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r30","arm":"g448","t":252,"nopts":4,"probs":[0.9998489086577421,5.411658523877711e-05,5.7382217473055725e-05,3.959253954591247e-05],"probs_oracle":[0.9998513460159302,5.3302719607017934e-05,5.6550619774498045e-05,3.8857629988342524e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.437358188056571e-06,"mean_abs_dp":1.2044327689858431e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9997947953961557,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r31","arm":"g256","t":113,"nopts":4,"probs":[0.008631585664368176,0.0011059898511099813,0.0003507446640268294,0.9899116798204951],"probs_oracle":[0.00852695107460022,0.0010953464079648256,0.0003487524518277496,0.9900289177894592],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00011723796896412342,"mean_abs_dp":5.86270535190787e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.981501966714859,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r31","arm":"g448","t":245,"nopts":4,"probs":[0.05102200279506944,0.000733497908383507,0.00045543920816355755,0.9477890600883835],"probs_oracle":[0.05128991976380348,0.000733718799892813,0.00045598449651151896,0.94752037525177],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0002686848366134642,"mean_abs_dp":0.00013434199630119398,"finite":true,"read_from":"main","oracle_top2_margin":0.8962304554879665,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r32","arm":"g256","t":104,"nopts":2,"probs":[0.007815888789288626,0.9921841112107114],"probs_oracle":[0.007742323447018862,0.9922577142715454],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.360306083403767e-05,"mean_abs_dp":7.358420155190084e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9845153908245265,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r32","arm":"g448","t":236,"nopts":2,"probs":[0.0037362205834114964,0.9962637794165885],"probs_oracle":[0.0036633117124438286,0.9963366985321045],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.291911551599028e-05,"mean_abs_dp":7.291399324182903e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9926733868196607,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g256","t":114,"nopts":4,"probs":[0.0008821606024939227,0.9981153486515879,0.0007313368263790275,0.0002711539195392441],"probs_oracle":[0.0008657005382701755,0.9981381893157959,0.000728585640899837,0.0002674604766070843],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.2840664207990358e-05,"mean_abs_dp":1.143633921077197e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9972724887775257,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g256","t":143,"nopts":2,"probs":[0.0045384317247953825,0.9954615682752046],"probs_oracle":[0.0045051840133965015,0.9954948425292969],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.32742540922748e-05,"mean_abs_dp":3.326098274557786e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9909896585159004,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g448","t":246,"nopts":4,"probs":[0.0005789839774674345,0.9988898631163828,0.00038871145382221865,0.00014244145232745353],"probs_oracle":[0.0005686153890565038,0.9989047050476074,0.00038564109127037227,0.00014100427506491542],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.4841931224585103e-05,"mean_abs_dp":7.429514862475082e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9983360896585509,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r33","arm":"g448","t":275,"nopts":2,"probs":[0.02017731832286999,0.97982268167713],"probs_oracle":[0.020169703289866447,0.979830265045166],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.615033003543353e-06,"mean_abs_dp":7.599200519755808e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9596605617552996,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r34","arm":"g256","t":120,"nopts":7,"probs":[0.9988085691426685,0.0005396284875608618,0.00019391955668154542,7.418084155665966e-05,9.053392721335287e-05,7.683514608511448e-05,0.00021633289823399582],"probs_oracle":[0.9988266825675964,0.0005331173888407648,0.000190935141290538,7.275486859725788e-05,8.913177589420229e-05,7.561183883808553e-05,0.00021174417634028941],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.811342492796797e-05,"mean_abs_dp":5.178441779765734e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9982935651787557,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r34","arm":"g448","t":252,"nopts":7,"probs":[0.998382774098152,0.0006926013096889132,0.00025281096181459895,0.0001009548384659366,0.00013218527314852875,0.00010375340967840937,0.00033492010905156595],"probs_oracle":[0.9983891248703003,0.0006920418818481266,0.0002512198407202959,0.00010025671508628875,0.00013134843902662396,0.00010331271914765239,0.000332709081703797],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.350772148278061e-06,"mean_abs_dp":1.8125709233494673e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9976970829884522,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g256","t":113,"nopts":4,"probs":[0.00023592110913123952,0.9995495093354037,0.00012432168817498734,9.02478672900214e-05],"probs_oracle":[0.00023222366871777922,0.9995563626289368,0.00012253291788510978,8.888279262464494e-05],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.85329353311559e-06,"mean_abs_dp":3.426144725457476e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999324138960219,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g256","t":144,"nopts":2,"probs":[0.9864678513445929,0.013532148655407106],"probs_oracle":[0.9866299629211426,0.013370047323405743],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00016211157654966168,"mean_abs_dp":0.00016210645427551258,"finite":true,"read_from":"main","oracle_top2_margin":0.9732599155977368,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g448","t":245,"nopts":4,"probs":[0.0002715254801668075,0.999483060267027,0.0001344148596126923,0.00011099939319346963],"probs_oracle":[0.0002710831176955253,0.9994843006134033,0.00013441758346743882,0.00011026520951418206],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2403463762744593e-06,"mean_abs_dp":6.049040953976947e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9992132174957078,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"r35","arm":"g448","t":276,"nopts":2,"probs":[0.9856076636305632,0.014392336369436931],"probs_oracle":[0.985750675201416,0.014249387197196484],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00014301157085283567,"mean_abs_dp":0.00014298037154664156,"finite":true,"read_from":"main","oracle_top2_margin":0.9715012880042195,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t01","arm":"text","t":61,"nopts":3,"probs":[0.0007730942012331925,0.9989533957382214,0.0002735100605455179],"probs_oracle":[0.0007738324929960072,0.9989522695541382,0.0002739227202255279],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.1261840832110792e-06,"mean_abs_dp":7.590451753452418e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9981784370611422,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":45,"nopts":2,"probs":[0.0009399438806628558,0.9990600561193372],"probs_oracle":[0.0009324414422735572,0.9990676045417786],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.548422441328029e-06,"mean_abs_dp":7.525430415313342e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.998135163099505,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":71,"nopts":2,"probs":[0.0037362205834114964,0.9962637794165885],"probs_oracle":[0.00368038984015584,0.9963196516036987],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.5872187110228566e-05,"mean_abs_dp":5.5851465182942504e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9926392617635429,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t02","arm":"text","t":106,"nopts":4,"probs":[0.9949650639523868,0.002838661462184466,0.0012304686645937255,0.0009658059208349796],"probs_oracle":[0.9950647950172424,0.002778948051854968,0.001205416745506227,0.0009508960647508502],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.973106485561267e-05,"mean_abs_dp":4.9851562589184664e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9922858469653875,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t03","arm":"text","t":50,"nopts":5,"probs":[0.004838391787132163,0.0021137368290455316,0.004616825695852084,0.9815003694468871,0.006930676241083267],"probs_oracle":[0.00478515587747097,0.0020656364504247904,0.004542955197393894,0.9817220568656921,0.006884180940687656],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00022168741880501663,"mean_abs_dp":8.867790118815034e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9748378759250045,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t04","arm":"text","t":46,"nopts":3,"probs":[0.016171076522082467,0.12137066942082927,0.8624582540570883],"probs_oracle":[0.016153892502188683,0.12061421573162079,0.863231897354126],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0007736432970376628,"mean_abs_dp":0.0005157603353799778,"finite":true,"read_from":"main","oracle_top2_margin":0.7426176816225052,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t04","arm":"text","t":73,"nopts":2,"probs":[0.9870795269609209,0.012920473039078976],"probs_oracle":[0.9872117638587952,0.012788254767656326],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0001322368978742272,"mean_abs_dp":0.00013222758464843856,"finite":true,"read_from":"main","oracle_top2_margin":0.9744235090911388,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t05","arm":"text","t":77,"nopts":7,"probs":[0.9990010092136432,0.0003148231548041622,0.0002080853990622202,8.914750147654019e-05,0.00011446765773008557,0.00014526702024785418,0.00012720005303579993],"probs_oracle":[0.9990049004554749,0.00031437192228622735,0.00020690326346084476,8.894488564692438e-05,0.00011396020272513852,0.00014418178761843592,0.00012674936442635953],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.89124183164391e-06,"mean_abs_dp":1.1100860034822475e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986905285331886,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t06","arm":"text","t":63,"nopts":4,"probs":[0.9948614098963506,0.0004318901268514135,0.00024132388922423246,0.004465376087573707],"probs_oracle":[0.9948415160179138,0.0004282791051082313,0.00024067162303254008,0.004489557351917028],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.4181264343321393e-05,"mean_abs_dp":1.208460767875307e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9903519586659968,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}},{"id":"t06","arm":"text","t":94,"nopts":3,"probs":[0.0009771764667452134,0.9988448664783218,0.00017795705493297366],"probs_oracle":[0.0009630977874621749,0.9988608360290527,0.00017604485037736595],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5969550730909887e-05,"mean_abs_dp":1.0653478189852036e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9978977382415906,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":true,"full_logits_max_abs_diff":0.0,"argmax_equal":true}}],"asset_identity":{"decoder_aimodelc_tree_sha256":"418213bfe687110f0169c95256951b0c7816f9d32521492c29c5a9503c396e25","decoder_equals_round6":true,"tower_g256_equals_round6":true,"tower_g448_equals_round6":true},"s1_same_pass":{"ids_equal_oracle":76,"slots_equal_oracle":76,"result":"PASS"},"note":"extra arm, not in the acceptance table: the alternative ship candidate (round 7 decides between the two bundles), same Swift code, same bar"},"s4":{"asset":"jit","bundle":"decider_2b_vision_decode_int8mix_pf16","fixture_json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/jit/fixture_jit.json","label":"S4 JIT .aimodel int8mix_pf16 (Swift runtime JIT, final binary, GPU shared with round 7)","assets":{"asset":"jit","bundle":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles/decider_2b_vision_decode_int8mix_pf16","decoder":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles/decider_2b_vision_decode_int8mix_pf16/decider_2b_vision_decode_int8mix_pf16.aimodel","decoder_options":"preferred gpu, allowed [\"cpu\", \"gpu\", \"neuralEngine\"], expectFrequentReshapes true","tower_options":"preferred gpu, allowed [\"cpu\", \"gpu\", \"neuralEngine\"], expectFrequentReshapes false"},"environment":{"argv":["/Users/majimadaisuke/code/coreai/_decider2bv/swift/.build/release/decider-vision","fixture","--bundle","/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles/decider_2b_vision_decode_int8mix_pf16","--asset","jit","--tower-g256","/Users/majimadaisuke/code/coreai/_decider2bv/exports/decider_2b_vision_g256_vision_fp16w32/decider_2b_vision_g256_vision_fp16w32.aimodel","--tower-g448","/Users/majimadaisuke/code/coreai/_decider2bv/exports/decider_2b_vision_g448_vision_fp16w32/decider_2b_vision_g448_vision_fp16w32.aimodel","--rows","/Users/majimadaisuke/code/coreai/_decider2bv/fixtures/rows.json","--images","/Users/majimadaisuke/code/coreai/_decider2bv/fixtures/images","--meta","/Users/majimadaisuke/code/coreai/_decider2bv/fixtures/meta.json","--out","/Users/majimadaisuke/code/coreai/_decider2bv/swift/jit/fixture_jit.json","--label","S4 JIT .aimodel int8mix_pf16 (Swift runtime JIT, final binary, GPU shared with round 7)"],"build_configuration":"Release","chip":"Apple M4 Max","device_architecture":"h16c","model":"Mac16,9","os":"Version 27.0 (Build 26A428)","os_build":"26A428","pid":83385},"load_first":{"decoder_s":{"main":3.697282333,"model":0.003015667,"prefill":3.4439825},"tokenizer_s":0.7415552910000001,"tower_s":{"g256":1.250896,"g448":1.199682958},"wall_s":10.340929333},"function_names":["main","prefill"],"reset_check":{"bit_equal":true,"run":"r01/g256","slot_logits_max_abs_diff":0,"wall_from_file_s":0.351541583},"bar":{"max_abs_dp":0.02,"mean_of_run_mean_abs_dp":0.002,"source":"readout_gate_vision.BAR (round 4, unchanged)","slots":"all","argmax":"all","full_vocab_top1":"the oracle's argmax letter on every slot","reset":"bit-equal re-run"},"summary":{"runs":76,"runs_expected":76,"slots_read":108,"slots_expected":108,"argmax_equal":108,"full_vocab_top1_is_oracle_letter":108,"max_abs_dp":0.010269998106143152,"worst":{"id":"r29","arm":"g256","t":163,"max_abs_dp":0.010269998106143152,"probs":[0.00762881364578451,0.8817709670392486,0.05378837403042393,0.056811845284543215],"probs_oracle":[0.007844753563404083,0.8715009689331055,0.05929643288254738,0.06135784834623337]},"mean_of_run_mean_abs_dp":0.0002628548906497969,"finite_all":true,"reset_bit_equal":true,"swift_json_probs_equal_recomputed":108,"slots_read_from_prefill":9},"checks":{"all_runs":true,"slots":true,"argmax":true,"full_vocab_top1":true,"max_abs_dp":true,"mean_of_run_mean_abs_dp":true,"finite":true,"reset_bit_equal":true},"result":"PASS","per_arm":{"g256":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.010269998106143152,"mean_of_run_mean_abs_dp":0.0002691914323512635},"g448":{"runs":35,"slots":49,"argmax_equal":49,"full_vocab_top1_is_oracle_letter":49,"max_abs_dp":0.008047910855851326,"mean_of_run_mean_abs_dp":0.0002601974476278511},"text":{"runs":6,"slots":10,"argmax_equal":10,"full_vocab_top1_is_oracle_letter":10,"max_abs_dp":0.0027482776324738722,"mean_of_run_mean_abs_dp":0.0002413934816859253}},"vs_python_same_bundle":{"definition":"per slot, the same bundle read by Python: image runs vs the round-6 e2e arm (decider_2b_vision_decode_int8mix_pf16+tower_fp16w32+chunk16: Pillow resize -> same tower .aimodelc -> same decoder .aimodelc, chunk order), text runs vs the round-6 b1 transcript (identical inputs). |dp| over the question's letters; bit-equality of the fp16 logits and of the tower's float32 output","all":{"slots":108,"max_abs_dp":0.0025122668999981723,"mean_abs_dp_of_slot_max":0.00011254457502275586,"letter_logits_bit_equal":30,"full_logits_bit_equal":0,"full_logits_max_abs_diff":0.015625,"argmax_equal":108,"tower_embeds_bit_equal_runs":68,"tower_embeds_max_abs_diff":0.053189992904663086},"by_arm":{"g256":{"slots":49,"max_abs_dp":0.0025122668999981723,"mean_abs_dp_of_slot_max":0.00012694501587716078,"letter_logits_bit_equal":14,"full_logits_bit_equal":0,"full_logits_max_abs_diff":0.015625,"argmax_equal":49,"tower_embeds_bit_equal_runs":35,"tower_embeds_max_abs_diff":0.0},"g448":{"slots":49,"max_abs_dp":0.001597164219744096,"mean_abs_dp_of_slot_max":9.85073638444937e-05,"letter_logits_bit_equal":13,"full_logits_bit_equal":0,"full_logits_max_abs_diff":0.015625,"argmax_equal":49,"tower_embeds_bit_equal_runs":33,"tower_embeds_max_abs_diff":0.053189992904663086},"text":{"slots":10,"max_abs_dp":0.0008402614305532546,"mean_abs_dp_of_slot_max":0.0001107647496096561,"letter_logits_bit_equal":3,"full_logits_bit_equal":0,"full_logits_max_abs_diff":0.015625,"argmax_equal":10,"tower_embeds_bit_equal_runs":0,"tower_embeds_max_abs_diff":0.0}},"note":"the Python references are the AOT .aimodelc runs (the Python runtime's JIT mis-executes hybrids above 0.8B), so this compares Swift's JIT specialization with the AOT compile of the same .aimodel: not expected bit-equal"},"python_bundle_records":{"e2e":"b75795b701f2f3c5fc5af4905819d59c73f2872fb0a92889713fed5295aedaa4","b1":"b75795b701f2f3c5fc5af4905819d59c73f2872fb0a92889713fed5295aedaa4"},"runs":[{"id":"r01","arm":"g256","tokens":105,"max_abs_dp":2.5768715975482692e-06,"mean_abs_dp":1.7408191129301828e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r01","arm":"g448","tokens":237,"max_abs_dp":3.350354069286965e-06,"mean_abs_dp":2.2244210755100423e-06,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g256","tokens":105,"max_abs_dp":1.1246624266325611e-06,"mean_abs_dp":7.339133635139662e-07,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r02","arm":"g448","tokens":237,"max_abs_dp":7.918799640725151e-07,"mean_abs_dp":5.289386100870975e-07,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g256","tokens":105,"max_abs_dp":1.011389407612473e-05,"mean_abs_dp":6.728917250436782e-06,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r03","arm":"g448","tokens":237,"max_abs_dp":1.5796380499533136e-05,"mean_abs_dp":1.0553039244136881e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g256","tokens":153,"max_abs_dp":2.5983379817710525e-05,"mean_abs_dp":1.223955231011191e-05,"calls":{"main":9,"prefill":9,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r04","arm":"g448","tokens":285,"max_abs_dp":1.182188369441839e-05,"mean_abs_dp":5.516199852741702e-06,"calls":{"main":13,"prefill":17,"total":30},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g256","tokens":105,"max_abs_dp":9.758937576465065e-05,"mean_abs_dp":6.507999532947386e-05,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r05","arm":"g448","tokens":237,"max_abs_dp":0.00011468986016494043,"mean_abs_dp":7.646138137069158e-05,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g256","tokens":126,"max_abs_dp":1.3670643266627813e-05,"mean_abs_dp":3.990369341271406e-06,"calls":{"main":14,"prefill":7,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r06","arm":"g448","tokens":258,"max_abs_dp":4.017709540959924e-05,"mean_abs_dp":1.0050060057700096e-05,"calls":{"main":2,"prefill":16,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g256","tokens":184,"max_abs_dp":0.000555964786662852,"mean_abs_dp":0.00010522828047692871,"calls":{"main":24,"prefill":10,"total":34},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r07","arm":"g448","tokens":316,"max_abs_dp":0.001313301258862587,"mean_abs_dp":0.0002375853556003894,"calls":{"main":28,"prefill":18,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g256","tokens":120,"max_abs_dp":2.3828163205230624e-05,"mean_abs_dp":7.938680486562087e-06,"calls":{"main":8,"prefill":7,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r08","arm":"g448","tokens":252,"max_abs_dp":5.35289815794301e-06,"mean_abs_dp":2.4908713158919714e-06,"calls":{"main":12,"prefill":15,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g256","tokens":128,"max_abs_dp":9.761840332157216e-05,"mean_abs_dp":2.5397773403390553e-05,"calls":{"main":0,"prefill":8,"total":8},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r09","arm":"g448","tokens":260,"max_abs_dp":0.0012878015402648657,"mean_abs_dp":0.00032691769202498845,"calls":{"main":4,"prefill":16,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g256","tokens":112,"max_abs_dp":5.449715737348981e-06,"mean_abs_dp":3.672967566215545e-06,"calls":{"main":0,"prefill":7,"total":7},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r10","arm":"g448","tokens":244,"max_abs_dp":1.605685906791443e-06,"mean_abs_dp":1.0810389389315276e-06,"calls":{"main":4,"prefill":15,"total":19},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g256","tokens":103,"max_abs_dp":1.5927995652420535e-06,"mean_abs_dp":1.564277811323114e-06,"calls":{"main":7,"prefill":6,"total":13},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r11","arm":"g448","tokens":235,"max_abs_dp":4.3411948593958735e-05,"mean_abs_dp":4.3387734207054295e-05,"calls":{"main":11,"prefill":14,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g256","tokens":117,"max_abs_dp":1.5215028256743146e-05,"mean_abs_dp":6.111018006804526e-06,"calls":{"main":5,"prefill":7,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r12","arm":"g448","tokens":249,"max_abs_dp":1.32729885379268e-05,"mean_abs_dp":5.309809711949323e-06,"calls":{"main":9,"prefill":15,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g256","tokens":140,"max_abs_dp":0.0010268573614620191,"mean_abs_dp":0.0002285656802879557,"calls":{"main":12,"prefill":8,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r13","arm":"g448","tokens":272,"max_abs_dp":2.347046169415279e-06,"mean_abs_dp":5.845733317851647e-07,"calls":{"main":0,"prefill":17,"total":17},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g256","tokens":111,"max_abs_dp":4.9575271670820875e-05,"mean_abs_dp":2.477187611122174e-05,"calls":{"main":15,"prefill":6,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r14","arm":"g448","tokens":243,"max_abs_dp":1.663801421147859e-05,"mean_abs_dp":8.32827485677639e-06,"calls":{"main":3,"prefill":15,"total":18},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g256","tokens":147,"max_abs_dp":6.0480405186025266e-05,"mean_abs_dp":2.7201462967334153e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r15","arm":"g448","tokens":279,"max_abs_dp":0.0002839418392824866,"mean_abs_dp":0.00011665318094784184,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g256","tokens":193,"max_abs_dp":7.170317945405383e-05,"mean_abs_dp":2.7272225783659654e-05,"calls":{"main":17,"prefill":11,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r16","arm":"g448","tokens":325,"max_abs_dp":9.344029149060962e-05,"mean_abs_dp":1.678207728538487e-05,"calls":{"main":21,"prefill":19,"total":40},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g256","tokens":171,"max_abs_dp":6.232178236716948e-06,"mean_abs_dp":2.3057802990526827e-06,"calls":{"main":27,"prefill":9,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r17","arm":"g448","tokens":303,"max_abs_dp":1.0346219082375558e-05,"mean_abs_dp":5.6775571041563825e-06,"calls":{"main":15,"prefill":18,"total":33},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g256","tokens":211,"max_abs_dp":0.0014163856175286142,"mean_abs_dp":0.0001750474030996274,"calls":{"main":35,"prefill":11,"total":46},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r18","arm":"g448","tokens":343,"max_abs_dp":0.0016786977299919625,"mean_abs_dp":0.00023020179123003936,"calls":{"main":23,"prefill":20,"total":43},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g256","tokens":177,"max_abs_dp":0.00043016399614692613,"mean_abs_dp":0.00011047142840621639,"calls":{"main":33,"prefill":9,"total":42},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r19","arm":"g448","tokens":309,"max_abs_dp":7.153999954923052e-05,"mean_abs_dp":2.522356569946268e-05,"calls":{"main":21,"prefill":18,"total":39},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.053189992904663086},{"id":"r20","arm":"g256","tokens":143,"max_abs_dp":0.0010516516420185917,"mean_abs_dp":0.00021033079697538394,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r20","arm":"g448","tokens":275,"max_abs_dp":0.0031754005345815406,"mean_abs_dp":0.0006350755143318498,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g256","tokens":143,"max_abs_dp":0.0003831379893546316,"mean_abs_dp":7.662552567816391e-05,"calls":{"main":15,"prefill":8,"total":23},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r21","arm":"g448","tokens":275,"max_abs_dp":0.0003551671702577419,"mean_abs_dp":7.103218840760138e-05,"calls":{"main":3,"prefill":17,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g256","tokens":147,"max_abs_dp":0.00012136867693079978,"mean_abs_dp":3.235720833762038e-05,"calls":{"main":3,"prefill":9,"total":12},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r22","arm":"g448","tokens":279,"max_abs_dp":0.0001342152059740087,"mean_abs_dp":3.314709721819102e-05,"calls":{"main":7,"prefill":17,"total":24},"tower_embeds_bit_equal_python":false,"tower_embeds_max_abs_diff_python":0.04187512397766113},{"id":"r23","arm":"g256","tokens":150,"max_abs_dp":7.761231113345701e-07,"mean_abs_dp":5.237824768687645e-07,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r23","arm":"g448","tokens":282,"max_abs_dp":2.75697468687941e-06,"mean_abs_dp":1.8197623077370293e-06,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g256","tokens":150,"max_abs_dp":3.1340847057026444e-05,"mean_abs_dp":2.089802593132409e-05,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r24","arm":"g448","tokens":282,"max_abs_dp":2.1980425674793052e-05,"mean_abs_dp":1.4640996968048604e-05,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g256","tokens":150,"max_abs_dp":0.00014589785278262823,"mean_abs_dp":9.72605009653385e-05,"calls":{"main":6,"prefill":9,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r25","arm":"g448","tokens":282,"max_abs_dp":0.00013002487058610956,"mean_abs_dp":8.668817530604759e-05,"calls":{"main":10,"prefill":17,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g256","tokens":145,"max_abs_dp":0.0002714486842951818,"mean_abs_dp":0.00014783695722291832,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r26","arm":"g448","tokens":277,"max_abs_dp":0.008047910855851326,"mean_abs_dp":0.004042321252790644,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g256","tokens":145,"max_abs_dp":0.0010564056541463307,"mean_abs_dp":0.0005281996838594962,"calls":{"main":1,"prefill":9,"total":10},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r27","arm":"g448","tokens":277,"max_abs_dp":0.0020540258285675073,"mean_abs_dp":0.0010270188514651482,"calls":{"main":5,"prefill":17,"total":22},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g256","tokens":170,"max_abs_dp":0.005176348202951542,"mean_abs_dp":0.0017342607152260844,"calls":{"main":10,"prefill":10,"total":20},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r28","arm":"g448","tokens":302,"max_abs_dp":0.0019455021340050727,"mean_abs_dp":0.0006484982763145303,"calls":{"main":14,"prefill":18,"total":32},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g256","tokens":164,"max_abs_dp":0.010269998106143152,"mean_abs_dp":0.005134999984394082,"calls":{"main":4,"prefill":10,"total":14},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r29","arm":"g448","tokens":296,"max_abs_dp":0.0009219564867415375,"mean_abs_dp":0.00048046172723646057,"calls":{"main":8,"prefill":18,"total":26},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g256","tokens":121,"max_abs_dp":2.0936016957850967e-06,"mean_abs_dp":1.3130068973401692e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r30","arm":"g448","tokens":253,"max_abs_dp":2.2734937715762626e-06,"mean_abs_dp":1.1509932108351578e-06,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g256","tokens":114,"max_abs_dp":0.000413855704184396,"mean_abs_dp":0.00020693592112921386,"calls":{"main":2,"prefill":7,"total":9},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r31","arm":"g448","tokens":246,"max_abs_dp":0.0012960828491945797,"mean_abs_dp":0.0006480418466028324,"calls":{"main":6,"prefill":15,"total":21},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g256","tokens":105,"max_abs_dp":0.00025746877992327555,"mean_abs_dp":0.00025744992064115953,"calls":{"main":9,"prefill":6,"total":15},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r32","arm":"g448","tokens":237,"max_abs_dp":0.0001315327521483889,"mean_abs_dp":0.00013152762987424216,"calls":{"main":13,"prefill":14,"total":27},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g256","tokens":144,"max_abs_dp":3.679974690538098e-05,"mean_abs_dp":1.58248809147756e-05,"calls":{"main":16,"prefill":8,"total":24},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r33","arm":"g448","tokens":276,"max_abs_dp":0.00014629366537866062,"mean_abs_dp":4.977629392031527e-05,"calls":{"main":20,"prefill":16,"total":36},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g256","tokens":121,"max_abs_dp":1.159026063290991e-05,"mean_abs_dp":3.408342709254556e-06,"calls":{"main":9,"prefill":7,"total":16},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r34","arm":"g448","tokens":253,"max_abs_dp":3.7819637253688754e-05,"mean_abs_dp":1.080754604865039e-05,"calls":{"main":13,"prefill":15,"total":28},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g256","tokens":145,"max_abs_dp":0.0003495572095629177,"mean_abs_dp":0.00011741245752116672,"calls":{"main":17,"prefill":8,"total":25},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"r35","arm":"g448","tokens":277,"max_abs_dp":0.00029367679892314605,"mean_abs_dp":9.934495250613614e-05,"calls":{"main":21,"prefill":16,"total":37},"tower_embeds_bit_equal_python":true,"tower_embeds_max_abs_diff_python":0.0},{"id":"t01","arm":"text","tokens":62,"max_abs_dp":7.073776392751974e-06,"mean_abs_dp":4.7075951419335e-06,"calls":{"main":14,"prefill":3,"total":17},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t02","arm":"text","tokens":107,"max_abs_dp":9.443508877182705e-05,"mean_abs_dp":5.232611016792211e-05,"calls":{"main":27,"prefill":5,"total":32},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t03","arm":"text","tokens":51,"max_abs_dp":0.00040389402743101765,"mean_abs_dp":0.0001615605446385609,"calls":{"main":3,"prefill":3,"total":6},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t04","arm":"text","tokens":74,"max_abs_dp":0.0027482776324738722,"mean_abs_dp":0.001164628404548942,"calls":{"main":26,"prefill":3,"total":29},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t05","arm":"text","tokens":78,"max_abs_dp":4.537688891637082e-06,"mean_abs_dp":1.9586454933179676e-06,"calls":{"main":14,"prefill":4,"total":18},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null},{"id":"t06","arm":"text","tokens":95,"max_abs_dp":0.00020233712833339634,"mean_abs_dp":6.31795901248751e-05,"calls":{"main":15,"prefill":5,"total":20},"tower_embeds_bit_equal_python":null,"tower_embeds_max_abs_diff_python":null}],"slots":[{"id":"r01","arm":"g256","t":104,"nopts":3,"probs":[0.999755396409714,0.0001365687873680299,0.00010803480291798109],"probs_oracle":[0.9997528195381165,0.00013744122406933457,0.0001098079519579187],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.5768715975482692e-06,"mean_abs_dp":1.7408191129301828e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996153783140471,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r01","arm":"g448","t":236,"nopts":3,"probs":[0.9996702795990644,0.00018090884456054606,0.0001488115563749729],"probs_oracle":[0.9996669292449951,0.00018169928807765245,0.00015134402201510966],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.350354069286965e-06,"mean_abs_dp":2.2244210755100423e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9994852299569175,"python":{"ref":"e2e","max_abs_dp":5.823449521355701e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r02","arm":"g256","t":104,"nopts":3,"probs":[0.00019560385992925842,0.9996434967937743,0.00016089934629637923],"probs_oracle":[0.00019598512153606862,0.9996423721313477,0.00016159516235347837],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.1246624266325611e-06,"mean_abs_dp":7.339133635139662e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9994463870098116,"python":{"ref":"e2e","max_abs_dp":6.271865670503076e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r02","arm":"g448","t":236,"nopts":3,"probs":[0.00026941041879441325,0.9994755169233607,0.0002550726578449945],"probs_oracle":[0.000269706710241735,0.9994760155677795,0.00025428077788092196],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.918799640725151e-07,"mean_abs_dp":5.289386100870975e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9992063088575378,"python":{"ref":"e2e","max_abs_dp":4.079435625192751e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r03","arm":"g256","t":104,"nopts":3,"probs":[0.00027150922489578924,0.00030526602705997904,0.9994232247480442],"probs_oracle":[0.00026771891862154007,0.0002989834756590426,0.9994333386421204],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.011389407612473e-05,"mean_abs_dp":6.728917250436782e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991343551664613,"python":{"ref":"e2e","max_abs_dp":2.112332341940341e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r03","arm":"g448","t":236,"nopts":3,"probs":[0.0004440067556332185,0.0004763497139775458,0.9990796435303891],"probs_oracle":[0.00043719500536099076,0.000467298727016896,0.9990954399108887],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5796380499533136e-05,"mean_abs_dp":1.0553039244136881e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9986281411838718,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r04","arm":"g256","t":112,"nopts":4,"probs":[0.9992577845074295,0.0004304228367887631,0.00023401683726508382,7.77758185169403e-05],"probs_oracle":[0.9992350339889526,0.0004433949070516974,0.0002398906071903184,8.168041676981375e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.275051847688836e-05,"mean_abs_dp":1.1375239229482671e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9987916390819009,"python":{"ref":"e2e","max_abs_dp":3.3751737166771794e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r04","arm":"g256","t":152,"nopts":5,"probs":[0.9963882543759199,0.00140724936659051,0.0009085884798557473,0.0008208412871331412,0.00047506649050058314],"probs_oracle":[0.996371865272522,0.0014012253377586603,0.0009149523684754968,0.0008468246669508517,0.0004651718772947788],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.5983379817710525e-05,"mean_abs_dp":1.2931002774615302e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9949706399347633,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r04","arm":"g448","t":244,"nopts":4,"probs":[0.9994427893016863,0.00030766628089341113,0.00016990945745593033,7.963495996452058e-05],"probs_oracle":[0.9994351267814636,0.00030979173607192934,0.0001716899569146335,8.336015162058175e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.662520222662117e-06,"mean_abs_dp":3.823416628986167e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9991253350453917,"python":{"ref":"e2e","max_abs_dp":2.4123083110231722e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r04","arm":"g448","t":284,"nopts":5,"probs":[0.9983800255236785,0.0006208420923134728,0.0004723117571141999,0.00032208863444249073,0.00020473199245146887],"probs_oracle":[0.9983682036399841,0.0006189235136844218,0.0004789685772266239,0.0003326288133393973,0.00020131732162553817],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.182188369441839e-05,"mean_abs_dp":6.870426431746131e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9977492801262997,"python":{"ref":"e2e","max_abs_dp":9.359735240410672e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r05","arm":"g256","t":104,"nopts":3,"probs":[0.00554799419077953,0.9932043718121355,0.0012476339970851306],"probs_oracle":[0.005630022846162319,0.9931067824363708,0.0012632559519261122],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.758937576465065e-05,"mean_abs_dp":6.507999532947386e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9874767595902085,"python":{"ref":"e2e","max_abs_dp":4.293715340442615e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r05","arm":"g448","t":236,"nopts":3,"probs":[0.00550069282476226,0.9924598441455958,0.002039463029641927],"probs_oracle":[0.005602917168289423,0.9923451542854309,0.0020519329700618982],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00011468986016494043,"mean_abs_dp":7.646138137069158e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9867422371171415,"python":{"ref":"e2e","max_abs_dp":5.82389347765222e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r06","arm":"g256","t":125,"nopts":8,"probs":[0.00044694069056577644,0.00042647378623073624,0.9978551480373706,0.00014173915085665984,0.00019373444914063905,0.00023552119337622425,0.0005967441144588498,0.00010369857800035812],"probs_oracle":[0.00044743184116669,0.00042624998604878783,0.997841477394104,0.0001445232192054391,0.00019250228069722652,0.00023652952222619206,0.0006083882763050497,0.00010282994480803609],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.3670643266627813e-05,"mean_abs_dp":3.990369341271406e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.997233089117799,"python":{"ref":"e2e","max_abs_dp":2.7492313426580317e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r06","arm":"g448","t":257,"nopts":8,"probs":[0.0004757458164931502,0.0004434438612828837,0.9978130495433771,0.00015205749602410513,0.000164413354757215,0.00023367849383437391,0.0005828959591108893,0.00013471547512019342],"probs_oracle":[0.00048085491289384663,0.0004459824995137751,0.9977728724479675,0.0001569857558934018,0.00016541295917704701,0.00023827303084544837,0.0006033360259607434,0.00013632865739054978],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.017709540959924e-05,"mean_abs_dp":1.0050060057700096e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9971695364220068,"python":{"ref":"e2e","max_abs_dp":6.27777399753171e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g256","t":112,"nopts":4,"probs":[0.9987773031434928,0.0005789187346181336,0.0002489898451253857,0.0003947882767636558],"probs_oracle":[0.9987712502479553,0.00058116385480389,0.0002529850753489882,0.000394599192077294],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.052895537456138e-06,"mean_abs_dp":3.120582658294209e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981900863931514,"python":{"ref":"e2e","max_abs_dp":4.502593486222493e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g256","t":156,"nopts":6,"probs":[0.0016061101786271905,0.9957564679004233,0.0005464514668958667,0.0005507373388693371,0.000989495776314906,0.0005507373388693371],"probs_oracle":[0.0015793738421052694,0.9958256483078003,0.0005403549876064062,0.0005451973993331194,0.0009696456836536527,0.0005397982313297689],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.918040737702302e-05,"mean_abs_dp":2.305706048757398e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.994246274465695,"python":{"ref":"e2e","max_abs_dp":4.317088933873821e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g256","t":183,"nopts":2,"probs":[0.06008665017400762,0.9399133498259924],"probs_oracle":[0.06064261496067047,0.9393573999404907],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.000555964786662852,"mean_abs_dp":0.000555957336082262,"finite":true,"read_from":"main","oracle_top2_margin":0.8787147849798203,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g448","t":244,"nopts":4,"probs":[0.9985704151618903,0.0005879135717614934,0.0002979400972307627,0.0005437311691173174],"probs_oracle":[0.9985532164573669,0.0005945708835497499,0.0003056767163798213,0.0005464947316795588],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.7198704523324615e-05,"mean_abs_dp":8.589049505720265e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9979586455738172,"python":{"ref":"e2e","max_abs_dp":8.794017005042676e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g448","t":288,"nopts":6,"probs":[0.0017091046515143318,0.9954124140695666,0.0006336757680525778,0.0006094000306012636,0.0010447547175155257,0.0005906507627496457],"probs_oracle":[0.0016743798041716218,0.9955074787139893,0.0006255566840991378,0.0006036655395291746,0.001013218774460256,0.0005756971659138799],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.506464442265994e-05,"mean_abs_dp":3.16887677803224e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9938330989098176,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r07","arm":"g448","t":315,"nopts":2,"probs":[0.07055959618729296,0.9294404038127071],"probs_oracle":[0.07187289744615555,0.9281271696090698],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.001313301258862587,"mean_abs_dp":0.0013132677312499286,"finite":true,"read_from":"main","oracle_top2_margin":0.8562542721629143,"python":{"ref":"e2e","max_abs_dp":0.0005140731928905962,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r08","arm":"g256","t":119,"nopts":6,"probs":[0.0006816103084406795,0.0005350020610329988,0.9980120199775113,0.0003853475287825401,0.00013526970486628972,0.00025075041936585156],"probs_oracle":[0.0006678571808151901,0.0005291247507557273,0.9980358481407166,0.0003851840447168797,0.00013258290709927678,0.00024942721938714385],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.3828163205230624e-05,"mean_abs_dp":7.938680486562087e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9973679909599014,"python":{"ref":"e2e","max_abs_dp":4.19381066500094e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r08","arm":"g448","t":251,"nopts":6,"probs":[0.0005609104497225779,0.00048353503167796504,0.9984269472004392,0.0002568035168493159,9.939407623030732e-05,0.00017240972508046788],"probs_oracle":[0.0005603667814284563,0.00047818213352002203,0.9984318614006042,0.00025924580404534936,9.779584797797725e-05,0.00017250367091037333],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.35289815794301e-06,"mean_abs_dp":2.4908713158919714e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9978714946191758,"python":{"ref":"e2e","max_abs_dp":2.4732396780580146e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r09","arm":"g256","t":127,"nopts":8,"probs":[0.0006933299429406053,0.0006933299429406053,0.0005571062601673248,0.017196196317700346,0.9762812540601513,0.003493619186395339,0.0007614739736621397,0.0003236903160422521],"probs_oracle":[0.0006885188049636781,0.0006795556400902569,0.0005544525920413435,0.01720018871128559,0.9762108325958252,0.0035912375897169113,0.0007585337734781206,0.0003167196991853416],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.761840332157216e-05,"mean_abs_dp":2.5397773403390553e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9590106438845396,"python":{"ref":"e2e","max_abs_dp":0.00013164016565887782,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r09","arm":"g448","t":259,"nopts":8,"probs":[0.001144253149969525,0.0009266444402602349,0.0006674369523453798,0.002137746321709652,0.9546212389830444,0.037304992072413175,0.0024413829301876983,0.0007563051500698042],"probs_oracle":[0.0011424915865063667,0.0009199993801303208,0.0006675715558230877,0.0021507099736481905,0.9533334374427795,0.03857970982789993,0.0024612408597022295,0.0007448457181453705],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0012878015402648657,"mean_abs_dp":0.00032691769202498845,"finite":true,"read_from":"main","oracle_top2_margin":0.9147537276148796,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r10","arm":"g256","t":111,"nopts":3,"probs":[0.9985617738331518,0.0011332232632495307,0.0003050029035985819],"probs_oracle":[0.9985672235488892,0.0011292636627331376,0.00030339331715367734],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.449715737348981e-06,"mean_abs_dp":3.672967566215545e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.997437959886156,"python":{"ref":"e2e","max_abs_dp":2.3728331185736374e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r10","arm":"g448","t":243,"nopts":3,"probs":[0.9995627366864256,0.00032754835696266016,0.00010971495661160856],"probs_oracle":[0.9995611310005188,0.0003284418198745698,0.00011045892460970208],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.605685906791443e-06,"mean_abs_dp":1.0810389389315276e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992326891806442,"python":{"ref":"e2e","max_abs_dp":6.882836893762345e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r11","arm":"g256","t":102,"nopts":2,"probs":[0.002822832897761561,0.9971771671022386],"probs_oracle":[0.002824368653818965,0.9971755743026733],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5927995652420535e-06,"mean_abs_dp":1.564277811323114e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9943512056488544,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r11","arm":"g448","t":234,"nopts":2,"probs":[0.007345826574477061,0.9926541734255229],"probs_oracle":[0.007302414625883102,0.992697536945343],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.3411948593958735e-05,"mean_abs_dp":4.3387734207054295e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9853951223194599,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r12","arm":"g256","t":116,"nopts":5,"probs":[0.9988417783865636,0.0006560432161776689,0.0003027142118159374,0.00011670672512478195,8.275746031800493e-05],"probs_oracle":[0.9988265633583069,0.0006655653123743832,0.0003061715397052467,0.00011903217091457918,8.272226841654629e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5215028256743146e-05,"mean_abs_dp":6.111018006804526e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981609980459325,"python":{"ref":"e2e","max_abs_dp":7.906999887286048e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r12","arm":"g448","t":248,"nopts":5,"probs":[0.9991691520092288,0.0004510392500393702,0.00020171722988287935,0.000101826777991372,7.626473285760427e-05],"probs_oracle":[0.9991558790206909,0.0004565727722365409,0.00020724281785078347,0.00010403238411527127,7.625338912475854e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.32729885379268e-05,"mean_abs_dp":5.309809711949323e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986993062484544,"python":{"ref":"e2e","max_abs_dp":2.973054286275128e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r13","arm":"g256","t":139,"nopts":9,"probs":[6.595857144707268e-05,0.00012322683446399792,9.522231728528029e-05,3.953981304927586e-05,0.00012226787563020434,0.00017108351679612007,0.00020636611313680245,0.016139530296771846,0.9830368046614193],"probs_oracle":[6.637163460254669e-05,0.00012154777505202219,9.527682414045557e-05,4.0484348573954776e-05,0.00012243450328242034,0.00017381930956616998,0.00021287811978254467,0.017157258465886116,0.9820099472999573],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010268573614620191,"mean_abs_dp":0.0002285656802879557,"finite":true,"read_from":"main","oracle_top2_margin":0.9648526888340712,"python":{"ref":"e2e","max_abs_dp":1.2750644822312474e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r13","arm":"g448","t":271,"nopts":9,"probs":[2.0298232978651756e-05,3.320569299158946e-05,3.320569299158946e-05,8.937191025708555e-06,2.7101696756657934e-05,3.1438518246070406e-05,4.3308388103081214e-05,9.168385833369109e-05,0.999710820728573],"probs_oracle":[2.0277637304388918e-05,3.2944568374659866e-05,3.3252264984184876e-05,9.089779268833809e-06,2.7153426344739273e-05,3.251702946727164e-05,4.40161456936039e-05,9.227909322362393e-05,0.9997084736824036],"probs_swift_json_equal":true,"argmax":8,"argmax_oracle":8,"argmax_equal":true,"full_vocab_top1_id":40,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.347046169415279e-06,"mean_abs_dp":5.845733317851647e-07,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9996161945891799,"python":{"ref":"e2e","max_abs_dp":3.8802258024439595e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r14","arm":"g256","t":110,"nopts":4,"probs":[0.9989638170624423,0.0006260772949370135,0.0001779782475088611,0.00023212739511180568],"probs_oracle":[0.9989142417907715,0.0006516185821965337,0.0001856259477790445,0.00024845064035616815],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.9575271670820875e-05,"mean_abs_dp":2.477187611122174e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.998262623208575,"python":{"ref":"e2e","max_abs_dp":1.8060136608602604e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r14","arm":"g448","t":242,"nopts":4,"probs":[0.9992736661170862,0.00045108642920291004,0.00011813348455712587,0.00015711396915362553],"probs_oracle":[0.9992570281028748,0.0004573110491037369,0.00012069327203789726,0.00016500464698765427],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.663801421147859e-05,"mean_abs_dp":8.32827485677639e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.998799717053771,"python":{"ref":"e2e","max_abs_dp":3.5088112734119417e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r15","arm":"g256","t":114,"nopts":4,"probs":[0.0013021585212991049,0.0009526795255484737,0.9968978312427982,0.0008473307103542585],"probs_oracle":[0.0013125522527843714,0.0009568282985128462,0.9968631267547607,0.00086748949252069],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.4704488037440306e-05,"mean_abs_dp":1.7351443663377714e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9955505745019764,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r15","arm":"g256","t":146,"nopts":3,"probs":[0.9973207895266072,0.0019709387912108783,0.0007082716821820267],"probs_oracle":[0.9973812699317932,0.001912115840241313,0.000706570572219789],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.0480405186025266e-05,"mean_abs_dp":4.0334822039276076e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9954691540915519,"python":{"ref":"e2e","max_abs_dp":5.551078335333569e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r15","arm":"g448","t":246,"nopts":4,"probs":[0.006571974705538923,0.0033045964115712466,0.9830174299221682,0.007105998960721564],"probs_oracle":[0.006657375022768974,0.0033173311967402697,0.9827334880828857,0.007291808724403381],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0002839418392824866,"mean_abs_dp":0.00014197167634084473,"finite":true,"read_from":"main","oracle_top2_margin":0.9754416793584824,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r15","arm":"g448","t":278,"nopts":3,"probs":[0.9946418468197052,0.004429648428586928,0.0009285047517077865],"probs_oracle":[0.9947662353515625,0.004311799071729183,0.0009220570791512728],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012438853185725574,"mean_abs_dp":8.289518709050465e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9904544362798333,"python":{"ref":"e2e","max_abs_dp":4.179758592115501e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r16","arm":"g256","t":117,"nopts":4,"probs":[0.0007721924615901811,0.0010804915133379933,0.9977882131808808,0.0003591028441909826],"probs_oracle":[0.0007772505050525069,0.0010821042815223336,0.997784435749054,0.0003561374032869935],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.058043462325811e-06,"mean_abs_dp":3.3534210943863635e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9967023314675316,"python":{"ref":"e2e","max_abs_dp":8.46517171535291e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r16","arm":"g256","t":165,"nopts":7,"probs":[0.0015807213866532268,0.003851668059482278,0.9877022803627443,0.0037919534810878382,0.0010779583298369502,0.0010366623947264762,0.0009587559854686503],"probs_oracle":[0.001566192484460771,0.0037890495732426643,0.9877481460571289,0.003863656660541892,0.0010514376917853951,0.0010348198702558875,0.0009466473711654544],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.170317945405383e-05,"mean_abs_dp":3.3598291299435215e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.983884489396587,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r16","arm":"g256","t":192,"nopts":2,"probs":[0.003649987031209837,0.9963500129687901],"probs_oracle":[0.0035970157478004694,0.9964029788970947],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.297128340936777e-05,"mean_abs_dp":5.296860585699177e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9928059631492943,"python":{"ref":"e2e","max_abs_dp":2.8301552534859553e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r16","arm":"g448","t":249,"nopts":4,"probs":[0.0001672454702239959,0.0004237527357745947,0.9992648271868377,0.00014417460716368048],"probs_oracle":[0.00016730134666431695,0.00042254829895682633,0.9992651343345642,0.00014504071441479027],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.2044368177683553e-06,"mean_abs_dp":6.083920589185248e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9988425860356074,"python":{"ref":"e2e","max_abs_dp":1.2141376643071133e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r16","arm":"g448","t":297,"nopts":7,"probs":[0.0012178399273033897,0.0026187673712410714,0.9924769003419017,0.0017174295718552749,0.0006123683723619794,0.0007048316761406575,0.0006518627391958755],"probs_oracle":[0.0012208223342895508,0.0025618993677198887,0.9925703406333923,0.0017182007431983948,0.0005935655790381134,0.0006946290959604084,0.0006406526081264019],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.344029149060962e-05,"mean_abs_dp":2.7753911130666015e-05,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9900084412656724,"python":{"ref":"e2e","max_abs_dp":9.546645649705175e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r16","arm":"g448","t":324,"nopts":2,"probs":[0.0044337652006436815,0.9955662347993564],"probs_oracle":[0.004423052072525024,0.9955769777297974],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0742930441010046e-05,"mean_abs_dp":1.0728029279833545e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9911539256572723,"python":{"ref":"e2e","max_abs_dp":3.461906584745211e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g256","t":108,"nopts":3,"probs":[0.999793305333327,0.00014090928886091258,6.578537781193891e-05],"probs_oracle":[0.9997918009757996,0.00014244449266698211,6.58148928778246e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.5352038060695327e-06,"mean_abs_dp":1.023025466464276e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996493564831326,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g256","t":141,"nopts":3,"probs":[0.9992366449824038,0.00039806762839613775,0.0003652873892001707],"probs_oracle":[0.999238133430481,0.00039801813545636833,0.0003638898197095841],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.4884480771248931e-06,"mean_abs_dp":9.785035024936258e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9988401152950246,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g256","t":170,"nopts":2,"probs":[0.9981469406043857,0.0018530593956144392],"probs_oracle":[0.9981531500816345,0.0018468272173777223],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":6.232178236716948e-06,"mean_abs_dp":6.220827742773878e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9963063228642568,"python":{"ref":"e2e","max_abs_dp":1.4394107925861863e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g448","t":240,"nopts":3,"probs":[0.9998114905071781,0.00012930798386508665,5.920150895686227e-05],"probs_oracle":[0.9998090863227844,0.0001311514643020928,5.978461922495626e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.404184393678399e-06,"mean_abs_dp":1.6102583662595095e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9996779348584823,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g448","t":273,"nopts":3,"probs":[0.999184405694298,0.000423718631874856,0.0003918756738270545],"probs_oracle":[0.9991744160652161,0.00042735174065455794,0.0003981590270996094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.989629081941231e-06,"mean_abs_dp":6.6353637113993474e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9987470643245615,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r17","arm":"g448","t":302,"nopts":2,"probs":[0.9979809860995514,0.0020190139004485808],"probs_oracle":[0.9979913234710693,0.002008667681366205],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.0346219082375558e-05,"mean_abs_dp":1.0341795300137244e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9959826557897031,"python":{"ref":"e2e","max_abs_dp":1.568061427216841e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r18","arm":"g256","t":123,"nopts":6,"probs":[0.9988553292715376,0.0006821862617323293,0.000168488118103511,7.05111942992542e-05,0.00012326854049633627,0.0001002166138306934],"probs_oracle":[0.9988755583763123,0.000659995072055608,0.00016870087711140513,7.21013275324367e-05,0.00012259691720828414,0.00010095535981236026],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.2191189676721234e-05,"mean_abs_dp":7.6055926603554994e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9982155633042566,"python":{"ref":"e2e","max_abs_dp":1.1288626380334676e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r18","arm":"g256","t":164,"nopts":5,"probs":[0.003239943825527225,0.015947562430114998,0.9488431368539634,0.028652580011214557,0.003316776879179632],"probs_oracle":[0.0031590552534908056,0.015553153119981289,0.9480063915252686,0.03006896562874317,0.0032123548444360495],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0014163856175286142,"mean_abs_dp":0.0005665701726274292,"finite":true,"read_from":"main","oracle_top2_margin":0.9179374258965254,"python":{"ref":"e2e","max_abs_dp":0.00022093236436164163,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.00909423828125,"argmax_equal":true}},{"id":"r18","arm":"g256","t":210,"nopts":6,"probs":[0.000534388468287204,0.0009233404731951682,0.0005470611256128158,0.000467926177544461,0.9968674020773639,0.0006598816779964693],"probs_oracle":[0.0005212401156313717,0.0009053464164026082,0.000543770962394774,0.0004741615557577461,0.9969097971916199,0.0006456233095377684],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.2395114255966426e-05,"mean_abs_dp":1.622023893239777e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9960044507752173,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r18","arm":"g448","t":255,"nopts":6,"probs":[0.998963180695481,0.00044743697991434504,0.00019854895626779276,8.50619190573578e-05,0.00014870629718051581,0.00015706515209919206],"probs_oracle":[0.9989650249481201,0.00044237051042728126,0.00019978830823674798,8.695968426764011e-05,0.0001475344179198146,0.0001582982367835939],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.066469487063777e-06,"mean_abs_dp":2.0754672084274686e-06,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9985226544376928,"python":{"ref":"e2e","max_abs_dp":9.139934626922752e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r18","arm":"g448","t":296,"nopts":5,"probs":[0.008432453443750518,0.045585412630546045,0.9227886870852241,0.018274840300098842,0.004918606540380547],"probs_oracle":[0.008088547736406326,0.04423723742365837,0.9244673848152161,0.018492545932531357,0.004714324604719877],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0016786977299919625,"mean_abs_dp":0.0007585532424634025,"finite":true,"read_from":"main","oracle_top2_margin":0.8802301473915577,"python":{"ref":"e2e","max_abs_dp":0.00013963682057124144,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r18","arm":"g448","t":342,"nopts":6,"probs":[0.0006442223699975897,0.0008805475778656093,0.0005729832370249538,0.0004900985344147757,0.9962903015629726,0.0011218467177244232],"probs_oracle":[0.0006307050934992731,0.0008567753247916698,0.0005710572004318237,0.0004954338655807078,0.9963390827178955,0.0011069673346355557],"probs_swift_json_equal":true,"argmax":4,"argmax_oracle":4,"argmax_equal":true,"full_vocab_top1_id":36,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.878115492290558e-05,"mean_abs_dp":1.803523922384855e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.99523211538326,"python":{"ref":"e2e","max_abs_dp":5.353280393460835e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r19","arm":"g256","t":110,"nopts":3,"probs":[0.9930769715055101,0.005813502185985843,0.001109526308503902],"probs_oracle":[0.9930190443992615,0.005865252576768398,0.0011157557601109147],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":5.7927106248634175e-05,"mean_abs_dp":3.863564954606744e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9871537918224931,"python":{"ref":"e2e","max_abs_dp":4.532872430449645e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g256","t":146,"nopts":4,"probs":[0.0025187323482134924,0.9848661449981463,0.011737845017151156,0.0008772776364891371],"probs_oracle":[0.0024147704243659973,0.9852963089942932,0.011448029428720474,0.0008408919093199074],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00043016399614692613,"mean_abs_dp":0.0002150818088985831,"finite":true,"read_from":"main","oracle_top2_margin":0.9738482795655727,"python":{"ref":"e2e","max_abs_dp":0.00020529517191569813,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g256","t":176,"nopts":2,"probs":[0.9979809860995514,0.0020190139004485808],"probs_oracle":[0.9979719519615173,0.0020279884338378906],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.034138034103023e-06,"mean_abs_dp":9.004335711706437e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9959439635276794,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":242,"nopts":3,"probs":[0.9957976485505746,0.0035634585384851643,0.0006388929109403547],"probs_oracle":[0.9957261085510254,0.0036251493729650974,0.0006488442304544151],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.153999954923052e-05,"mean_abs_dp":4.772738451440803e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9921009591780603,"python":{"ref":"e2e","max_abs_dp":5.007672684093122e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.009765625,"argmax_equal":true}},{"id":"r19","arm":"g448","t":278,"nopts":4,"probs":[0.0007311464943971777,0.9978555869855932,0.0011867679022261688,0.00022649861778363273],"probs_oracle":[0.000710506399627775,0.9978955984115601,0.0011723366333171725,0.00022145592083688825],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.0011425966879344e-05,"mean_abs_dp":2.0031371648005717e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9967232617782429,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r19","arm":"g448","t":308,"nopts":2,"probs":[0.9981469406043857,0.0018530593956144392],"probs_oracle":[0.9981451034545898,0.0018549266969785094],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.867301364070215e-06,"mean_abs_dp":1.8522255799585706e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9962901767576113,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r20","arm":"g256","t":142,"nopts":10,"probs":[0.00037366873502835634,0.001015736932290859,0.0004300900776693149,0.0005653427416987194,0.0006713611170990793,0.00034022925496708894,0.0014102110735206232,0.4679806227681372,0.001047979883644456,0.5261647574159444],"probs_oracle":[0.0003781993000302464,0.0010252433130517602,0.00043770880438387394,0.0005875966744497418,0.0006762796547263861,0.00034810963552445173,0.0014829656574875116,0.46888846158981323,0.001062334282323718,0.5251131057739258],"probs_swift_json_equal":true,"argmax":9,"argmax_oracle":9,"argmax_equal":true,"full_vocab_top1_id":41,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010516516420185917,"mean_abs_dp":0.00021033079697538394,"finite":true,"read_from":"main","oracle_top2_margin":0.05622464418411255,"python":{"ref":"e2e","max_abs_dp":0.0019422593624780493,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r20","arm":"g448","t":274,"nopts":10,"probs":[0.00044629102898849994,0.0010622650751165128,0.0006147918594341929,0.0006595753841167259,0.0009596765211833384,0.00041275172850905304,0.0020001376268076886,0.8931713927182668,0.0012419136940403494,0.09943120436353692],"probs_oracle":[0.0004485415993258357,0.0010800594463944435,0.0006290642777457833,0.0006908925715833902,0.000972151174210012,0.00042305333772674203,0.0021294124890118837,0.8899959921836853,0.0012687055859714746,0.10236208140850067],"probs_swift_json_equal":true,"argmax":7,"argmax_oracle":7,"argmax_equal":true,"full_vocab_top1_id":39,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0031754005345815406,"mean_abs_dp":0.0006350755143318498,"finite":true,"read_from":"main","oracle_top2_margin":0.7876339107751846,"python":{"ref":"e2e","max_abs_dp":8.32252869794269e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r21","arm":"g256","t":142,"nopts":10,"probs":[0.00038066532774827446,0.0006786093185264187,0.0003836509229169286,0.9706008427623039,0.0009275489637960032,0.00048498020915693253,0.0005162587453323793,0.0007280415234258034,0.00042466286030250146,0.024874739366491247],"probs_oracle":[0.00038343475898727775,0.000686735613271594,0.0003942575422115624,0.9702177047729492,0.0009427570621483028,0.0004960547666996717,0.0005290430854074657,0.0007387973600998521,0.0004333727993071079,0.02517782151699066],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003831379893546316,"mean_abs_dp":7.662552567816391e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9450398832559586,"python":{"ref":"e2e","max_abs_dp":5.318758712392099e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r21","arm":"g448","t":274,"nopts":10,"probs":[0.0005377889309338181,0.0008797639752014322,0.0005336038242703335,0.9647783467967226,0.0012119260640375607,0.0006692891382811684,0.0006692891382811684,0.0010047219961689022,0.0005814884803318632,0.029133781655771202],"probs_oracle":[0.0005481093539856374,0.0008933480712585151,0.0005559168639592826,0.9644231796264648,0.0012373479548841715,0.0006897316197864711,0.0006905003683641553,0.0010131854796782136,0.0005959380650892854,0.02935273014008999],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003551671702577419,"mean_abs_dp":7.103218840760138e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9350704494863749,"python":{"ref":"e2e","max_abs_dp":4.215644724038818e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r22","arm":"g256","t":146,"nopts":9,"probs":[0.0014978381908845161,0.0019383466404100241,0.0009823062678760053,0.9885214703386129,0.001751150258340247,0.0021122904128913706,0.0015095858814151535,0.0008402110762125343,0.0008468009333572036],"probs_oracle":[0.0015125663485378027,0.0019855706486850977,0.0009997023735195398,0.9884001016616821,0.0017508834134787321,0.0021429003681987524,0.0015369588509202003,0.0008162282756529748,0.0008550662896595895],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00012136867693079978,"mean_abs_dp":3.235720833762038e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9862572012934834,"python":{"ref":"e2e","max_abs_dp":1.5172942463387209e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r22","arm":"g448","t":278,"nopts":9,"probs":[0.002461689884941835,0.0036954342057595397,0.001577017743278318,0.9853880856931871,0.0015048008429878572,0.0016918930656287052,0.0017869952162547634,0.0006943515151015031,0.001199731832860278],"probs_oracle":[0.0024558305740356445,0.0037613804452121258,0.0016030978877097368,0.9852538704872131,0.0015132761327549815,0.0017176602268591523,0.0018075669649988413,0.0006852896767668426,0.0012020787689834833],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0001342152059740087,"mean_abs_dp":3.314709721819102e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.981492490042001,"python":{"ref":"e2e","max_abs_dp":4.7217250608744266e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.01171875,"argmax_equal":true}},{"id":"r23","arm":"g256","t":149,"nopts":3,"probs":[0.9997602713814426,1.8054515155315643e-05,0.00022167410340184044],"probs_oracle":[0.9997594952583313,1.821711521188263e-05,0.00022230672766454518],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.761231113345701e-07,"mean_abs_dp":5.237824768687645e-07,"finite":true,"read_from":"main","oracle_top2_margin":0.9995371885306668,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r23","arm":"g448","t":281,"nopts":3,"probs":[0.9997180255865948,2.2822081264177322e-05,0.00025915233214096606],"probs_oracle":[0.999715268611908,2.3231894374475814e-05,0.00026144483126699924],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.75697468687941e-06,"mean_abs_dp":1.8197623077370293e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999453823780641,"python":{"ref":"e2e","max_abs_dp":2.121277215771933e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r24","arm":"g256","t":149,"nopts":3,"probs":[3.92579615874165e-05,0.9991752989387929,0.0007854430996196572],"probs_oracle":[3.9324964745901525e-05,0.9991439580917358,0.000816729327198118],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.1340847057026444e-05,"mean_abs_dp":2.089802593132409e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9983272287645377,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r24","arm":"g448","t":281,"nopts":3,"probs":[3.819318047770046e-05,0.9990218988383057,0.0009399079812165827],"probs_oracle":[3.7778951082145795e-05,0.9990003705024719,0.0009618884068913758],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.1980425674793052e-05,"mean_abs_dp":1.4640996968048604e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9980384820955805,"python":{"ref":"e2e","max_abs_dp":1.4889543460780175e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r25","arm":"g256","t":149,"nopts":3,"probs":[0.0012334548764589078,0.0013868105948845334,0.9973797345286566],"probs_oracle":[0.001173934550024569,0.0013004472712054849,0.9975256323814392],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00014589785278262823,"mean_abs_dp":9.72605009653385e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9962251851102337,"python":{"ref":"e2e","max_abs_dp":2.033805237655617e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r25","arm":"g448","t":281,"nopts":3,"probs":[0.002628406563880659,0.0012415713477821624,0.9961300220883372],"probs_oracle":[0.0025366705376654863,0.0012032677186653018,0.9962600469589233],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00013002487058610956,"mean_abs_dp":8.668817530604759e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9937233764212579,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r26","arm":"g256","t":144,"nopts":4,"probs":[0.9864878343376653,0.0012986671811051793,0.010873607178588872,0.001339891302640712],"probs_oracle":[0.986732542514801,0.001274453941732645,0.01060215849429369,0.0013908690307289362],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0002714486842951818,"mean_abs_dp":0.00014783695722291832,"finite":true,"read_from":"main","oracle_top2_margin":0.9761303840205073,"python":{"ref":"e2e","max_abs_dp":2.0414421091241586e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r26","arm":"g448","t":276,"nopts":4,"probs":[0.8412513458000042,0.029932743475269523,0.08593923534134366,0.04287667538338271],"probs_oracle":[0.8332034349441528,0.030570058152079582,0.08590248227119446,0.050323981791734695],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.008047910855851326,"mean_abs_dp":0.004042321252790644,"finite":true,"read_from":"main","oracle_top2_margin":0.7473009526729584,"python":{"ref":"e2e","max_abs_dp":0.001597164219744096,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r27","arm":"g256","t":144,"nopts":4,"probs":[0.0078269701772855,0.008731616201164524,0.005254779190857521,0.9781866344306925],"probs_oracle":[0.007379894610494375,0.008370750583708286,0.005006327293813229,0.9792430400848389],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0010564056541463307,"mean_abs_dp":0.0005281996838594962,"finite":true,"read_from":"main","oracle_top2_margin":0.9708722895011306,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r27","arm":"g448","t":276,"nopts":4,"probs":[0.011080446893041284,0.014006997326846068,0.008170209195044827,0.9667423465850677],"probs_oracle":[0.010312587022781372,0.013162228278815746,0.007728788536041975,0.9687963724136353],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0020540258285675073,"mean_abs_dp":0.0010270188514651482,"finite":true,"read_from":"main","oracle_top2_margin":0.9556341441348195,"python":{"ref":"e2e","max_abs_dp":0.00018877593540722604,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r28","arm":"g256","t":169,"nopts":6,"probs":[0.028324282480161406,0.7960460014278966,0.17215702130306876,0.0012444821192614934,0.0008290038123807796,0.0013992088572311006],"probs_oracle":[0.029139939695596695,0.7908696532249451,0.17654110491275787,0.0012475387193262577,0.0008099511615000665,0.0013918428448960185],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.005176348202951542,"mean_abs_dp":0.0017342607152260844,"finite":true,"read_from":"main","oracle_top2_margin":0.6143285483121872,"python":{"ref":"e2e","max_abs_dp":0.0025122668999981723,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.01171875,"argmax_equal":true}},{"id":"r28","arm":"g448","t":301,"nopts":6,"probs":[0.02419895823249143,0.7411363259323706,0.23139404597695684,0.00125278877212783,0.0007748415934970148,0.0012430394925562573],"probs_oracle":[0.023014351725578308,0.7404738664627075,0.23333954811096191,0.0012221261858940125,0.000736995309125632,0.001213126815855503],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0019455021340050727,"mean_abs_dp":0.0006484982763145303,"finite":true,"read_from":"main","oracle_top2_margin":0.5071343183517456,"python":{"ref":"e2e","max_abs_dp":0.0014960277477021888,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r29","arm":"g256","t":163,"nopts":4,"probs":[0.00762881364578451,0.8817709670392486,0.05378837403042393,0.056811845284543215],"probs_oracle":[0.007844753563404083,0.8715009689331055,0.05929643288254738,0.06135784834623337],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.010269998106143152,"mean_abs_dp":0.005134999984394082,"finite":true,"read_from":"main","oracle_top2_margin":0.8101431205868721,"python":{"ref":"e2e","max_abs_dp":0.0007642277728411306,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.009765625,"argmax_equal":true}},{"id":"r29","arm":"g448","t":295,"nopts":4,"probs":[0.0024837397052332303,0.9788005270433822,0.015944921384408648,0.002770811866976021],"probs_oracle":[0.002556135645136237,0.9778785705566406,0.016833467409014702,0.002731863409280777],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0009219564867415375,"mean_abs_dp":0.00048046172723646057,"finite":true,"read_from":"main","oracle_top2_margin":0.9610451031476259,"python":{"ref":"e2e","max_abs_dp":0.00014258011857115616,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.01171875,"argmax_equal":true}},{"id":"r30","arm":"g256","t":120,"nopts":4,"probs":[0.9998231008534353,6.787565261550054e-05,6.013449537365e-05,4.888899857578681e-05],"probs_oracle":[0.9998210072517395,6.738237425452098e-05,6.102487168391235e-05,5.0663769798120484e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.0936016957850967e-06,"mean_abs_dp":1.3130068973401692e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999753624877485,"python":{"ref":"e2e","max_abs_dp":2.6462011306606305e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r30","arm":"g448","t":252,"nopts":4,"probs":[0.9998536195097018,5.327783637121997e-05,5.5617014284713345e-05,3.748563964216082e-05],"probs_oracle":[0.9998513460159302,5.3302719607017934e-05,5.6550619774498045e-05,3.8857629988342524e-05],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.2734937715762626e-06,"mean_abs_dp":1.1509932108351578e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9997947953961557,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r31","arm":"g256","t":113,"nopts":4,"probs":[0.008902913131610247,0.0011230700387340725,0.00035895474438093475,0.9896150620852748],"probs_oracle":[0.00852695107460022,0.0010953464079648256,0.0003487524518277496,0.9900289177894592],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.000413855704184396,"mean_abs_dp":0.00020693592112921386,"finite":true,"read_from":"main","oracle_top2_margin":0.981501966714859,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r31","arm":"g448","t":245,"nopts":4,"probs":[0.052554705495051535,0.0007555322105085357,0.0004654698918644934,0.9462242924025754],"probs_oracle":[0.05128991976380348,0.000733718799892813,0.00045598449651151896,0.94752037525177],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0012960828491945797,"mean_abs_dp":0.0006480418466028324,"finite":true,"read_from":"main","oracle_top2_margin":0.8962304554879665,"python":{"ref":"e2e","max_abs_dp":3.6490129841686364e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r32","arm":"g256","t":104,"nopts":2,"probs":[0.007999754508377905,0.9920002454916221],"probs_oracle":[0.007742323447018862,0.9922577142715454],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00025746877992327555,"mean_abs_dp":0.00025744992064115953,"finite":true,"read_from":"main","oracle_top2_margin":0.9845153908245265,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r32","arm":"g448","t":236,"nopts":2,"probs":[0.003794834220043924,0.9962051657799561],"probs_oracle":[0.0036633117124438286,0.9963366985321045],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0001315327521483889,"mean_abs_dp":0.00013152762987424216,"finite":true,"read_from":"main","oracle_top2_margin":0.9926733868196607,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r33","arm":"g256","t":114,"nopts":4,"probs":[0.0008550443852981514,0.9981445457419723,0.0007313582196052213,0.0002690516531245098],"probs_oracle":[0.0008657005382701755,0.9981381893157959,0.000728585640899837,0.0002674604766070843],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.06561529720241e-05,"mean_abs_dp":5.344083592811623e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9972724887775257,"python":{"ref":"e2e","max_abs_dp":6.700417199975413e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r33","arm":"g256","t":143,"nopts":2,"probs":[0.0044683842664911206,0.9955316157335089],"probs_oracle":[0.0045051840133965015,0.9954948425292969],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.679974690538098e-05,"mean_abs_dp":3.678647555870355e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9909896585159004,"python":{"ref":"e2e","max_abs_dp":3.461906584745211e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r33","arm":"g448","t":246,"nopts":4,"probs":[0.0005655814386908304,0.998906836993382,0.00038569303116014277,0.000141888536766913],"probs_oracle":[0.0005686153890565038,0.9989047050476074,0.00038564109127037227,0.00014100427506491542],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.033950365673361e-06,"mean_abs_dp":1.525524433006878e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9983360896585509,"python":{"ref":"e2e","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r33","arm":"g448","t":275,"nopts":2,"probs":[0.020023441289455244,0.9799765587105447],"probs_oracle":[0.020169703289866447,0.979830265045166],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00014629366537866062,"mean_abs_dp":0.00014627783289493206,"finite":true,"read_from":"main","oracle_top2_margin":0.9596605617552996,"python":{"ref":"e2e","max_abs_dp":0.00015387703341474684,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r34","arm":"g256","t":120,"nopts":7,"probs":[0.9988382728282293,0.000531278122583029,0.00018795886871535687,7.106300695838998e-05,8.948183604486197e-05,7.389383533514724e-05,0.00020805150213380026],"probs_oracle":[0.9988266825675964,0.0005331173888407648,0.000190935141290538,7.275486859725788e-05,8.913177589420229e-05,7.561183883808553e-05,0.00021174417634028941],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.159026063290991e-05,"mean_abs_dp":3.408342709254556e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9982935651787557,"python":{"ref":"e2e","max_abs_dp":1.2003630549872746e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r34","arm":"g448","t":252,"nopts":7,"probs":[0.998426944507554,0.00068189368863174,0.000245043626961034,9.67130903392945e-05,0.0001296343135500781,0.00010017363336174561,0.0003195971396020285],"probs_oracle":[0.9983891248703003,0.0006920418818481266,0.0002512198407202959,0.00010025671508628875,0.00013134843902662396,0.00010331271914765239,0.000332709081703797],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":3.7819637253688754e-05,"mean_abs_dp":1.080754604865039e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9976970829884522,"python":{"ref":"e2e","max_abs_dp":8.473190525437602e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r35","arm":"g256","t":113,"nopts":4,"probs":[0.0002340861402635545,0.9995537100550135,0.00012335472753200293,8.884907719106707e-05],"probs_oracle":[0.00023222366871777922,0.9995563626289368,0.00012253291788510978,8.888279262464494e-05],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.6525739232896584e-06,"mean_abs_dp":1.3426426373839897e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.999324138960219,"python":{"ref":"e2e","max_abs_dp":3.477144381611397e-07,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r35","arm":"g256","t":144,"nopts":2,"probs":[0.9869795098861571,0.013020490113842825],"probs_oracle":[0.9866299629211426,0.013370047323405743],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0003495572095629177,"mean_abs_dp":0.00034955208728873217,"finite":true,"read_from":"main","oracle_top2_margin":0.9732599155977368,"python":{"ref":"e2e","max_abs_dp":0.00010078101270816209,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"r35","arm":"g448","t":245,"nopts":4,"probs":[0.00027365420963342555,0.9994798780934616,0.0001354686571135128,0.00011099903979144729],"probs_oracle":[0.0002710831176955253,0.9994843006134033,0.00013441758346743882,0.00011026520951418206],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.422519941704728e-06,"mean_abs_dp":2.194628950736049e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9992132174957078,"python":{"ref":"e2e","max_abs_dp":4.902173015630673e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"r35","arm":"g448","t":276,"nopts":2,"probs":[0.9860442896017267,0.013955710398273338],"probs_oracle":[0.985750675201416,0.014249387197196484],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00029367679892314605,"mean_abs_dp":0.00029364559961693633,"finite":true,"read_from":"main","oracle_top2_margin":0.9715012880042195,"python":{"ref":"e2e","max_abs_dp":0.00021338990237096747,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"t01","arm":"text","t":61,"nopts":3,"probs":[0.0007791512584474113,0.9989451957777454,0.0002756529638071723],"probs_oracle":[0.0007738324929960072,0.9989522695541382,0.0002739227202255279],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":7.073776392751974e-06,"mean_abs_dp":4.7075951419335e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9981784370611422,"python":{"ref":"b1","max_abs_dp":6.058715626050147e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t02","arm":"text","t":45,"nopts":2,"probs":[0.0009622125595974061,0.9990377874404025],"probs_oracle":[0.0009324414422735572,0.9990676045417786],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":2.981710137606175e-05,"mean_abs_dp":2.979410934995534e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.998135163099505,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t02","arm":"text","t":71,"nopts":2,"probs":[0.0037654137846738736,0.9962345862153261],"probs_oracle":[0.00368038984015584,0.9963196516036987],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":8.506538837260713e-05,"mean_abs_dp":8.504466644532042e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9926392617635429,"python":{"ref":"b1","max_abs_dp":2.9193201262378565e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t02","arm":"text","t":106,"nopts":4,"probs":[0.9949703599284706,0.0028386765717433424,0.0012401259508170363,0.0009508375489690355],"probs_oracle":[0.9950647950172424,0.002778948051854968,0.001205416745506227,0.0009508960647508502],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":9.443508877182705e-05,"mean_abs_dp":4.7232832438206336e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9922858469653875,"python":{"ref":"b1","max_abs_dp":4.672555108653498e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t03","arm":"text","t":50,"nopts":5,"probs":[0.004952211438149302,0.00212991959886443,0.004615968624190325,0.9813181628382611,0.006983737500535041],"probs_oracle":[0.00478515587747097,0.0020656364504247904,0.004542955197393894,0.9817220568656921,0.006884180940687656],"probs_swift_json_equal":true,"argmax":3,"argmax_oracle":3,"argmax_equal":true,"full_vocab_top1_id":35,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00040389402743101765,"mean_abs_dp":0.0001615605446385609,"finite":true,"read_from":"main","oracle_top2_margin":0.9748378759250045,"python":{"ref":"b1","max_abs_dp":3.834904269097144e-05,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t04","arm":"text","t":46,"nopts":3,"probs":[0.016516660185437806,0.12299972009291008,0.8604836197216521],"probs_oracle":[0.016153892502188683,0.12061421573162079,0.863231897354126],"probs_swift_json_equal":true,"argmax":2,"argmax_oracle":2,"argmax_equal":true,"full_vocab_top1_id":34,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.0027482776324738722,"mean_abs_dp":0.001832183225670761,"finite":true,"read_from":"main","oracle_top2_margin":0.7426176816225052,"python":{"ref":"b1","max_abs_dp":0.0008402614305532546,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t04","arm":"text","t":73,"nopts":2,"probs":[0.9873750507184357,0.012624949281564409],"probs_oracle":[0.9872117638587952,0.012788254767656326],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00016330548609191746,"mean_abs_dp":0.0001632961728662138,"finite":true,"read_from":"main","oracle_top2_margin":0.9744235090911388,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t05","arm":"text","t":77,"nopts":7,"probs":[0.9990094381443665,0.0003148258110803196,0.00020326682100547426,8.742397033439731e-05,0.00011581794344657404,0.0001430160705162348,0.0001262112392503973],"probs_oracle":[0.9990049004554749,0.00031437192228622735,0.00020690326346084476,8.894488564692438e-05,0.00011396020272513852,0.00014418178761843592,0.00012674936442635953],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":4.537688891637082e-06,"mean_abs_dp":1.9586454933179676e-06,"finite":true,"read_from":"main","oracle_top2_margin":0.9986905285331886,"python":{"ref":"b1","max_abs_dp":4.401235192741382e-06,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}},{"id":"t06","arm":"text","t":63,"nopts":4,"probs":[0.9950438531462472,0.0004252722655703612,0.0002357768575025097,0.0042950977306799865],"probs_oracle":[0.9948415160179138,0.0004282791051082313,0.00024067162303254008,0.004489557351917028],"probs_swift_json_equal":true,"argmax":0,"argmax_oracle":0,"argmax_equal":true,"full_vocab_top1_id":32,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":0.00020233712833339634,"mean_abs_dp":0.0001011745886595847,"finite":true,"read_from":"prefill","oracle_top2_margin":0.9903519586659968,"python":{"ref":"b1","max_abs_dp":0.00014265831968462983,"letter_logits_bit_equal":false,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.015625,"argmax_equal":true}},{"id":"t06","arm":"text","t":94,"nopts":3,"probs":[0.0009771737282980766,0.9988420673074887,0.00018075896421321802],"probs_oracle":[0.0009630977874621749,0.9988608360290527,0.00017604485037736595],"probs_swift_json_equal":true,"argmax":1,"argmax_oracle":1,"argmax_equal":true,"full_vocab_top1_id":33,"full_vocab_top1_is_oracle_letter":true,"full_vocab_top1_equals_swift_json":true,"max_abs_dp":1.8768721564033086e-05,"mean_abs_dp":1.2519592078595622e-05,"finite":true,"read_from":"main","oracle_top2_margin":0.9978977382415906,"python":{"ref":"b1","max_abs_dp":0.0,"letter_logits_bit_equal":true,"full_logits_bit_equal":false,"full_logits_max_abs_diff":0.0078125,"argmax_equal":true}}],"s1_same_pass":{"ids_equal_oracle":76,"slots_equal_oracle":76,"result":"PASS"},"note":"decoder .aimodel specialized with SpecializationOptions(preferredComputeUnitKind: .gpu) + expectFrequentReshapes (the exporter's AOT flags / the zoo engines' dynamic-model options), towers .aimodel with preferredComputeUnitKind .gpu. The first JIT pass of this round (swift/jit/fixture_jit_prelim.json, 02:49, same source bar two cosmetic lines, same verdict: 108/108 argmax, max|dp| 0.010270) created the CLI's Core AI cache entries for the .aimodel assets (~/Library/Caches/coreai-cache/26A428/decider-vision, first written by this round at 02:44): its load was the cold specialization, 14.35 s (decoder AIModel 9.42 s, main 0.38 s, prefill 0.52 s, towers 1.72 / 1.63 s); this pass loaded from that cache in 10.34 s. MPSGraph scratch stayed at 96 KB in both passes (swift/jit/monitor*.tsv); footprint after the runs 0.83 GB"},"generated_at":"2026-09-28T18:13:11+00:00","result_line":"S1 PASS | S2 PASS | S3 PASS | S4 PASS | C done (values side by side) | D done | S3_ALT PASS | D_CPU_SHARED done","c":{"image":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/c_check/circles3_300.png","image_sha256":"135a2cd21abfdc5aa4b9a807d33ca3787ab6ee1c6b5ca8d3e3f9ca9b9f91852d","grid":8,"context":"This is a visual question about the image.","questions":[{"options":["1","2","3","4","5"],"text":"How many circles are in the image?"},{"options":["red","green","blue","black"],"text":"What color are the circles?"}],"swift_json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/c_check/ask_swift.json","rows":[{"question":"How many circles are in the image?","option":"1","letter":"A","swift":0.0002980455024984349,"python_resize_bicubic":0.0002980455024984349,"python_pillow":0.0002980455024984349},{"question":"How many circles are in the image?","option":"2","letter":"B","swift":0.00040106406105301466,"python_resize_bicubic":0.00040106406105301466,"python_pillow":0.00040106406105301466},{"question":"How many circles are in the image?","option":"3","letter":"C","swift":0.9989236894706457,"python_resize_bicubic":0.9989236894706457,"python_pillow":0.9989236894706457},{"question":"How many circles are in the image?","option":"4","letter":"D","swift":0.00030750648233751756,"python_resize_bicubic":0.00030750648233751756,"python_pillow":0.00030750648233751756},{"question":"How many circles are in the image?","option":"5","letter":"E","swift":6.969448346543244e-05,"python_resize_bicubic":6.969448346543244e-05,"python_pillow":6.969448346543244e-05},{"question":"What color are the circles?","option":"red","letter":"A","swift":0.9994275694761531,"python_resize_bicubic":0.9994275694761531,"python_pillow":0.9994275694761531},{"question":"What color are the circles?","option":"green","letter":"B","swift":0.0002049468062048371,"python_resize_bicubic":0.0002049468062048371,"python_pillow":0.0002049468062048371},{"question":"What color are the circles?","option":"blue","letter":"C","swift":0.00027794908892031885,"python_resize_bicubic":0.00027794908892031885,"python_pillow":0.00027794908892031885},{"question":"What color are the circles?","option":"black","letter":"D","swift":8.953462872158028e-05,"python_resize_bicubic":8.953462872158028e-05,"python_pillow":8.953462872158028e-05}],"ids_equal":true,"slots_swift":[124,160],"slots_python":[124,160],"patches_sha256_equal_python_resize_bicubic":true,"tower_embeds_sha256_equal_python_resize_bicubic":true,"letter_logits":{"swift":[[8.5703125,8.8671875,16.6875,8.6015625,7.1171875],[16.859375,8.3671875,8.671875,7.5390625]],"python_resize_bicubic":[[8.5703125,8.8671875,16.6875,8.6015625,7.1171875],[16.859375,8.3671875,8.671875,7.5390625]],"python_pillow":[[8.5703125,8.8671875,16.6875,8.6015625,7.1171875],[16.859375,8.3671875,8.671875,7.5390625]]},"max_abs_dp_swift_vs_python_resize_bicubic":0.0,"max_abs_dp_swift_vs_python_pillow":0.0,"swift_wall_s":0.7477903330000001,"decoder":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/bundles_aotc/decider_2b_vision_decode_int8mix_pf16.h16c.aimodelc","tower":"/Users/majimadaisuke/code/coreai/_decider2bv/exports/decider_2b_vision_g256_vision_fp16w32_aotc/decider_2b_vision_g256_vision_fp16w32.h16c.aimodelc","result":"done (values side by side)"},"d":{"passes":{"int8mix_pf16_pass1":{"pass":1,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d2/int8mix_pf16_pass1.json","label":"D timing r8-d2 decider_2b_vision_decode_int8mix_pf16 pass 1 (lock taken)","bundle":"decider_2b_vision_decode_int8mix_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.470762416},"g256":{"wall_from_file_s":{"n":34,"median":0.32227904150000003,"min":0.203976916,"max":0.713300166},"decision_wall_s":{"n":34,"median":0.3220629585,"min":0.203762416,"max":0.713019708},"decoder_s":{"n":34,"median":0.29651458350000004,"min":0.17581991700000002,"max":0.682527375},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":21.967813,"min":21.461292,"max":23.636667},"host_preprocess_ms":{"n":34,"median":2.7012915,"min":1.368083,"max":5.553167000000001}},"g448":{"wall_from_file_s":{"n":35,"median":0.5839214170000001,"min":0.48461820800000005,"max":0.8492037500000001},"decision_wall_s":{"n":35,"median":0.5836782500000001,"min":0.48436729200000006,"max":0.8489620000000001},"decoder_s":{"n":35,"median":0.512474875,"min":0.41482970900000005,"max":0.777726542},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.348958,"min":60.60450000000001,"max":66.00179200000001},"host_preprocess_ms":{"n":35,"median":5.969209000000002,"min":4.860666,"max":8.039209000000001}},"text":{"wall_from_file_s":{"n":6,"median":0.283874938,"min":0.134275041,"max":0.443800167},"decision_wall_s":{"n":6,"median":0.283874938,"min":0.134275041,"max":0.443800167},"decoder_s":{"n":6,"median":0.281781125,"min":0.132884834,"max":0.44101204200000005},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.184875,"min":24.44675,"max":26.724375000000002},"main_call_ms":{"n":850,"median":11.563187500000002,"min":11.186667000000002,"max":34.673792},"state_reset_ms":{"n":75,"median":0.5337080000000001,"min":0.43054200000000004,"max":0.649459},"footprint_mb":{"after_load":426.2402,"after_runs":835.856896,"max_during_runs":1008.036352}},"int8mix_pf16_pass2":{"pass":2,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d2/int8mix_pf16_pass2.json","label":"D timing r8-d2 decider_2b_vision_decode_int8mix_pf16 pass 2 (lock taken)","bundle":"decider_2b_vision_decode_int8mix_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.46828320900000003},"g256":{"wall_from_file_s":{"n":34,"median":0.32091735400000004,"min":0.20458599900000002,"max":0.714281334},"decision_wall_s":{"n":34,"median":0.320671354,"min":0.204348083,"max":0.714014459},"decoder_s":{"n":34,"median":0.294870729,"min":0.176371458,"max":0.683382916},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":22.019271000000003,"min":21.623167000000002,"max":22.329250000000002},"host_preprocess_ms":{"n":34,"median":2.6823335,"min":1.3029580000000003,"max":5.388583}},"g448":{"wall_from_file_s":{"n":35,"median":0.578912916,"min":0.481927417,"max":0.8632179180000001},"decision_wall_s":{"n":35,"median":0.578715541,"min":0.481689625,"max":0.8629885840000001},"decoder_s":{"n":35,"median":0.5094761250000001,"min":0.41367979200000005,"max":0.7907798330000001},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.211125,"min":60.322250000000004,"max":64.98100000000001},"host_preprocess_ms":{"n":35,"median":6.101792,"min":5.489750000000001,"max":8.532792}},"text":{"wall_from_file_s":{"n":6,"median":0.285721771,"min":0.11204941700000001,"max":0.441940708},"decision_wall_s":{"n":6,"median":0.285721771,"min":0.11204941700000001,"max":0.441940708},"decoder_s":{"n":6,"median":0.2838064585,"min":0.11056,"max":0.438861125},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.144834000000003,"min":24.454417000000003,"max":28.713083},"main_call_ms":{"n":850,"median":11.587146,"min":11.087042,"max":15.988125000000002},"state_reset_ms":{"n":75,"median":0.533625,"min":0.42991700000000005,"max":0.6466250000000001},"footprint_mb":{"after_load":428.84528,"after_runs":841.01788,"max_during_runs":1013.230104}}},"loads":[{"pass":"int8mix_pf16_pass1","first_load_in_process":{"decoder_s":{"main":3.663784292,"model":0.002302291,"prefill":8.636227625},"tokenizer_s":0.6098415410000001,"tower_s":{"g256":0.128831459,"g448":0.104685333},"wall_s":13.1494765},"reload_in_process":{"decoder_s":{"main":0.38505587500000005,"model":0.000835291,"prefill":0.968548791},"tokenizer_s":0.615567791,"tower_s":{"g256":0.099190125,"g448":0.101095083},"wall_s":2.174112958}},{"pass":"int8mix_pf16_pass2","first_load_in_process":{"decoder_s":{"main":0.436315084,"model":0.0008599160000000001,"prefill":1.030513209},"tokenizer_s":0.638049708,"tower_s":{"g256":0.13008729200000002,"g448":0.103020416},"wall_s":2.342754417},"reload_in_process":{"decoder_s":{"main":0.40606508300000005,"model":0.00083575,"prefill":1.05394625},"tokenizer_s":0.624299958,"tower_s":{"g256":0.097684166,"g448":0.1023425},"wall_s":2.288885833}}],"gpu_lock":{"lock":"/Users/majimadaisuke/code/coreai/_GPU_LOCK","started":"2026-09-29 03:10:59 JST","load_avg_start":[3.9208984375,4.908203125,5.00341796875],"snapshot_before_lock":{"at":"before lock","time":"03:11:04","load_avg":[3.6064453125,4.82666015625,4.9736328125],"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:10:57","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":28,"etime":"01:10:56","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":24.1,"rss_mb":80,"etime":"14-12:33:46","command":"WindowServer"},{"pid":1258,"ppid":1,"pri":54,"cpu_pct":10.6,"rss_mb":21,"etime":"14-12:33:11","command":"WallpaperAerialsExtension"},{"pid":1786,"ppid":1,"pri":4,"cpu_pct":18.2,"rss_mb":276,"etime":"14-12:31:06","command":"cmux"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:45:36","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:39:07","command":"MTLCompilerService"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:33:33","command":"MTLCompilerService"},{"pid":30296,"ppid":1,"pri":4,"cpu_pct":76.7,"rss_mb":23,"etime":"07-04:55:39","command":"Storage"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":1.0,"rss_mb":320,"etime":"03-14:05:38","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"33:33","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"29:11","command":"swift-plugin-server"},{"pid":55391,"ppid":1,"pri":58,"cpu_pct":102.6,"rss_mb":970,"etime":"05-14:03:51","command":"mediaanalysisd"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:23:57","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:30:40","command":"com.apple.dt.SKAgent"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:34:55","command":"taskgated"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:47:14","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"06:35","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"06:35","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:44:11","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:44:10","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:44:09","command":"SWBBuildService"},{"pid":98712,"ppid":98711,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"00:06","command":"Python"},{"pid":98715,"ppid":98712,"pri":20,"cpu_pct":0.0,"rss_mb":25,"etime":"00:06","command":"Python"},{"pid":98718,"ppid":98715,"pri":20,"cpu_pct":0.0,"rss_mb":64,"etime":"00:05","command":"Python"},{"pid":98862,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"00:00","command":"zsh"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:46:30","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":0.1,"rss_mb":426,"etime":"06:10:11","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:10:10","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":5.9,"rss_mb":545,"etime":"04:52:18","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:52:17","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":4.2,"rss_mb":587,"etime":"04:52:18","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:52:16","command":"python"},{"pid":96617,"ppid":96397,"pri":31,"cpu_pct":3.1,"rss_mb":575,"etime":"01:12","command":"claude"},{"pid":96692,"ppid":96617,"pri":31,"cpu_pct":0.0,"rss_mb":61,"etime":"01:09","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":0.4,"rss_mb":386,"etime":"01:35:52","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:35:51","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:33:25","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":2.1,"rss_mb":657,"etime":"01:32:30","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:32:29","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":10.3,"rss_mb":595,"etime":"43:19","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"43:17","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"33:34","command":"sourcekit-lsp"}]},"file_existed":true,"file_bytes_before":0,"file_mtime_before":"2026-09-29 03:10:24","state":"taken by this run (flock; file 0 bytes, last written 2026-09-29 03:10:24; GPU utilization [0, 0, 0, 0, 0])","waited_s":4.2,"readings":[{"t_s":4.2,"flock_free":true,"tag":"","tag_pid_live":null,"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false}],"snapshot_after_lock":{"at":"after lock, before the passes","time":"03:11:12","load_avg":[3.27880859375,4.71630859375,4.9326171875],"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:11:06","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":28,"etime":"01:11:05","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":17.7,"rss_mb":80,"etime":"14-12:33:55","command":"WindowServer"},{"pid":1258,"ppid":1,"pri":54,"cpu_pct":10.4,"rss_mb":21,"etime":"14-12:33:20","command":"WallpaperAerialsExtension"},{"pid":1786,"ppid":1,"pri":4,"cpu_pct":5.8,"rss_mb":276,"etime":"14-12:31:15","command":"cmux"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:45:45","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:39:16","command":"MTLCompilerService"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:33:42","command":"MTLCompilerService"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":0.8,"rss_mb":321,"etime":"03-14:05:47","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"33:42","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"29:20","command":"swift-plugin-server"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:24:06","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:30:49","command":"com.apple.dt.SKAgent"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:35:04","command":"taskgated"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:47:23","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"06:44","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"06:44","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:44:20","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:44:19","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:44:18","command":"SWBBuildService"},{"pid":98712,"ppid":98711,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"00:15","command":"Python"},{"pid":98715,"ppid":98712,"pri":20,"cpu_pct":0.0,"rss_mb":25,"etime":"00:15","command":"Python"},{"pid":98718,"ppid":98715,"pri":20,"cpu_pct":0.0,"rss_mb":64,"etime":"00:14","command":"Python"},{"pid":98862,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"00:09","command":"zsh"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:46:39","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":2.7,"rss_mb":426,"etime":"06:10:20","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:10:19","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":0.3,"rss_mb":545,"etime":"04:52:27","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:52:26","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":0.9,"rss_mb":587,"etime":"04:52:27","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:52:25","command":"python"},{"pid":96617,"ppid":96397,"pri":31,"cpu_pct":2.7,"rss_mb":576,"etime":"01:21","command":"claude"},{"pid":96692,"ppid":96617,"pri":31,"cpu_pct":0.0,"rss_mb":61,"etime":"01:18","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":0.7,"rss_mb":386,"etime":"01:36:01","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:36:00","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:33:34","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":5.9,"rss_mb":657,"etime":"01:32:39","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:32:38","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":4.9,"rss_mb":595,"etime":"43:28","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"43:26","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"33:43","command":"sourcekit-lsp"}]},"passes":[{"bundle":"decider_2b_vision_decode_int8mix_pf16","pass":1,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d2/int8mix_pf16_pass1.json","exit":0,"s":55.3,"priority_every_5s":[[0.0,31,3.28,0],[5.0,31,3.26,0],[10.1,46,3.24,0],[15.1,31,3.06,86],[20.1,31,2.97,89],[25.1,31,3.13,90],[30.2,31,3.12,88],[35.2,31,3.27,86],[40.2,31,3.41,89],[45.3,31,3.46,87],[50.3,31,3.42,85]],"priority_min":31,"demoted_samples":0},{"bundle":"decider_2b_vision_decode_int8mix_pf16","pass":2,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d2/int8mix_pf16_pass2.json","exit":0,"s":45.3,"priority_every_5s":[[0.0,31,4.35,0],[5.0,31,4.32,87],[10.1,31,4.45,89],[15.1,31,4.42,88],[20.1,31,4.62,86],[25.1,31,4.81,88],[30.2,31,5.79,91],[35.2,31,5.41,89],[40.2,31,5.13,84]],"priority_min":31,"demoted_samples":0}],"snapshot_after_passes":{"at":"after the passes, lock still held","time":"03:12:57","load_avg":[4.8779296875,4.806640625,4.93359375],"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:12:51","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":28,"etime":"01:12:50","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":22.1,"rss_mb":80,"etime":"14-12:35:40","command":"WindowServer"},{"pid":1258,"ppid":1,"pri":54,"cpu_pct":8.0,"rss_mb":21,"etime":"14-12:35:05","command":"WallpaperAerialsExtension"},{"pid":1786,"ppid":1,"pri":4,"cpu_pct":6.3,"rss_mb":277,"etime":"14-12:33:00","command":"cmux"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:47:30","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:41:01","command":"MTLCompilerService"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:35:27","command":"MTLCompilerService"},{"pid":30296,"ppid":1,"pri":4,"cpu_pct":90.3,"rss_mb":23,"etime":"07-04:57:33","command":"Storage"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":0.9,"rss_mb":321,"etime":"03-14:07:32","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"35:27","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"31:05","command":"swift-plugin-server"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:25:51","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:32:34","command":"com.apple.dt.SKAgent"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:36:49","command":"taskgated"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:49:08","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"08:29","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"08:29","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:46:05","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:46:04","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:46:03","command":"SWBBuildService"},{"pid":98712,"ppid":98711,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"02:00","command":"Python"},{"pid":98715,"ppid":98712,"pri":20,"cpu_pct":0.0,"rss_mb":25,"etime":"02:00","command":"Python"},{"pid":98718,"ppid":98715,"pri":20,"cpu_pct":0.0,"rss_mb":64,"etime":"01:59","command":"Python"},{"pid":98862,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"01:54","command":"zsh"},{"pid":40412,"ppid":40051,"pri":31,"cpu_pct":5.7,"rss_mb":570,"etime":"08:48:25","command":"claude"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:48:24","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":5.0,"rss_mb":426,"etime":"06:12:05","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:12:04","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":4.6,"rss_mb":568,"etime":"04:54:12","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:54:11","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":8.4,"rss_mb":587,"etime":"04:54:12","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:54:10","command":"python"},{"pid":96617,"ppid":96397,"pri":31,"cpu_pct":3.1,"rss_mb":598,"etime":"03:06","command":"claude"},{"pid":96692,"ppid":96617,"pri":31,"cpu_pct":0.0,"rss_mb":61,"etime":"03:03","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":10.4,"rss_mb":387,"etime":"01:37:46","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:37:45","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:35:19","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":0.3,"rss_mb":657,"etime":"01:34:24","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:34:23","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":7.7,"rss_mb":596,"etime":"45:13","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"45:11","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"35:28","command":"sourcekit-lsp"}]},"load_avg_end":[4.8779296875,4.806640625,4.93359375],"finished":"2026-09-29 03:12:57 JST","file_after":"0 bytes"},"contended":false,"definition":"wall_from_file = ImageIO decode of the PNG + RGB8 + resize + patches + tower + prompt + static inputs + decoder calls (chunk order) + read-out, per run (one row = one decision with 1..3 questions); the first run of each process is reported apart (first-call costs); load = AIModel(contentsOf:) + loadFunction, tokenizer apart","result":"done","note":"headline: GPU lock taken, GPU 0% at the take, no other user process above half a core during the passes (gpu_lock.json snapshots); the earlier run r8-d (section d_cpu_shared) had another session's Python job (pid 92919, ~5 cores, 8 GB) on the CPU"},"d_cpu_shared":{"passes":{"int8mix_pf16_pass1":{"pass":1,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/int8mix_pf16_pass1.json","label":"D timing r8-d decider_2b_vision_decode_int8mix_pf16 pass 1 (lock taken)","bundle":"decider_2b_vision_decode_int8mix_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.532022709},"g256":{"wall_from_file_s":{"n":34,"median":0.326077021,"min":0.20581412500000001,"max":0.745955958},"decision_wall_s":{"n":34,"median":0.32551045850000004,"min":0.205263834,"max":0.7449246660000001},"decoder_s":{"n":34,"median":0.2994362085,"min":0.17663779200000002,"max":0.7141166670000001},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":21.980312,"min":21.304583,"max":22.576458000000002},"host_preprocess_ms":{"n":34,"median":2.8050415,"min":1.4976260000000001,"max":6.574084000000001}},"g448":{"wall_from_file_s":{"n":35,"median":0.590675583,"min":0.48979358300000003,"max":0.8835404590000001},"decision_wall_s":{"n":35,"median":0.590402167,"min":0.48951912500000005,"max":0.8832897500000001},"decoder_s":{"n":35,"median":0.5211739590000001,"min":0.420536334,"max":0.8103899170000001},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.22329200000001,"min":60.677125000000004,"max":68.4325},"host_preprocess_ms":{"n":35,"median":6.388416,"min":4.959208,"max":8.871834000000002}},"text":{"wall_from_file_s":{"n":6,"median":0.287513229,"min":0.119995208,"max":0.45156791700000004},"decision_wall_s":{"n":6,"median":0.287513229,"min":0.119995208,"max":0.45156791700000004},"decoder_s":{"n":6,"median":0.2854577495,"min":0.118508209,"max":0.44872441700000004},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.252166000000003,"min":24.255667000000003,"max":32.44200000000001},"main_call_ms":{"n":850,"median":12.0696665,"min":11.115166,"max":19.873125},"state_reset_ms":{"n":75,"median":0.5695,"min":0.432375,"max":0.664959},"footprint_mb":{"after_load":433.875168,"after_runs":841.722368,"max_during_runs":1014.196736}},"int8mix_pf16_pass2":{"pass":2,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/int8mix_pf16_pass2.json","label":"D timing r8-d decider_2b_vision_decode_int8mix_pf16 pass 2 (lock taken)","bundle":"decider_2b_vision_decode_int8mix_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.484434499},"g256":{"wall_from_file_s":{"n":34,"median":0.32087525000000006,"min":0.203918125,"max":0.7116941660000001},"decision_wall_s":{"n":34,"median":0.3206138545,"min":0.203682417,"max":0.7114523330000001},"decoder_s":{"n":34,"median":0.295073438,"min":0.17591075,"max":0.681835042},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":21.917291499999997,"min":21.239375000000003,"max":22.341208},"host_preprocess_ms":{"n":34,"median":2.3968540000000003,"min":1.260333,"max":4.796208}},"g448":{"wall_from_file_s":{"n":35,"median":0.582190333,"min":0.47856350000000003,"max":0.8574511240000001},"decision_wall_s":{"n":35,"median":0.5819572080000001,"min":0.47833275000000003,"max":0.8572099160000001},"decoder_s":{"n":35,"median":0.5107727080000001,"min":0.41145058300000004,"max":0.785944292},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.063958,"min":60.598917,"max":65.16591700000001},"host_preprocess_ms":{"n":35,"median":5.636417000000002,"min":5.149375,"max":7.725125000000001}},"text":{"wall_from_file_s":{"n":6,"median":0.290923562,"min":0.11478412500000001,"max":0.45685050000000005},"decision_wall_s":{"n":6,"median":0.290923562,"min":0.11478412500000001,"max":0.45685050000000005},"decoder_s":{"n":6,"median":0.28885220850000004,"min":0.113254458,"max":0.453969958},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.188042,"min":24.288583000000003,"max":43.34637500000001},"main_call_ms":{"n":850,"median":11.571896000000002,"min":11.118708000000002,"max":15.074750000000002},"state_reset_ms":{"n":75,"median":0.480208,"min":0.43979200000000007,"max":0.5974160000000001},"footprint_mb":{"after_load":434.67796,"after_runs":840.067608,"max_during_runs":1012.96796}},"fp16_pf16_pass1":{"pass":3,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/fp16_pf16_pass1.json","label":"D timing r8-d decider_2b_vision_decode_fp16_pf16 pass 1 (lock taken)","bundle":"decider_2b_vision_decode_fp16_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.617887791},"g256":{"wall_from_file_s":{"n":34,"median":0.3527769375,"min":0.21049258300000004,"max":0.7198905420000001},"decision_wall_s":{"n":34,"median":0.352066792,"min":0.20936125000000003,"max":0.7187059170000001},"decoder_s":{"n":34,"median":0.3252796875,"min":0.180599375,"max":0.6870614580000001},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":22.186604000000003,"min":21.374333000000004,"max":30.244292},"host_preprocess_ms":{"n":34,"median":2.7834375000000002,"min":1.3879580000000002,"max":5.801041}},"g448":{"wall_from_file_s":{"n":35,"median":0.6199610000000001,"min":0.48541579100000004,"max":1.039417667},"decision_wall_s":{"n":35,"median":0.6197294170000001,"min":0.485174541,"max":1.039172542},"decoder_s":{"n":35,"median":0.550469541,"min":0.416629291,"max":0.958596167},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.310500000000005,"min":60.881417000000006,"max":72.840375},"host_preprocess_ms":{"n":35,"median":5.778499,"min":5.133708,"max":7.649834000000001}},"text":{"wall_from_file_s":{"n":6,"median":0.284288229,"min":0.11197333300000001,"max":0.439407625},"decision_wall_s":{"n":6,"median":0.284288229,"min":0.11197333300000001,"max":0.439407625},"decoder_s":{"n":6,"median":0.2822564375,"min":0.11052558300000001,"max":0.436433},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.449125000000002,"min":24.599625000000003,"max":55.033792000000005},"main_call_ms":{"n":850,"median":11.655896000000002,"min":11.182209,"max":18.691042000000003},"state_reset_ms":{"n":75,"median":0.515583,"min":0.43025,"max":2.4232090000000004},"footprint_mb":{"after_load":428.977432,"after_runs":840.445544,"max_during_runs":1012.657768}},"fp16_pf16_pass2":{"pass":4,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/fp16_pf16_pass2.json","label":"D timing r8-d decider_2b_vision_decode_fp16_pf16 pass 2 (lock taken)","bundle":"decider_2b_vision_decode_fp16_pf16","asset":"aot","build":"Release","first_decision_of_process":{"id":"r01","arm":"g256","wall_from_file_s":0.461432416},"g256":{"wall_from_file_s":{"n":34,"median":0.32961558350000003,"min":0.204545207,"max":0.7049392080000001},"decision_wall_s":{"n":34,"median":0.32932102100000005,"min":0.20429666600000002,"max":0.7046855830000001},"decoder_s":{"n":34,"median":0.3034236665,"min":0.176299,"max":0.675305667},"tokens_median":143.5,"slots":48,"tower_ms":{"n":34,"median":21.889896,"min":21.232292,"max":26.225916},"host_preprocess_ms":{"n":34,"median":2.6034585000000003,"min":1.221291,"max":5.195959}},"g448":{"wall_from_file_s":{"n":35,"median":0.5880738340000001,"min":0.49495129200000004,"max":0.856885208},"decision_wall_s":{"n":35,"median":0.5878322920000001,"min":0.49471437500000004,"max":0.856645375},"decoder_s":{"n":35,"median":0.5177554590000001,"min":0.42606745900000004,"max":0.785339},"tokens_median":275.0,"slots":49,"tower_ms":{"n":35,"median":61.14704100000001,"min":60.556833000000005,"max":65.483917},"host_preprocess_ms":{"n":35,"median":5.844834,"min":4.753666000000001,"max":13.060041}},"text":{"wall_from_file_s":{"n":6,"median":0.28323675000000004,"min":0.110371833,"max":0.440906167},"decision_wall_s":{"n":6,"median":0.28323675000000004,"min":0.110371833,"max":0.440906167},"decoder_s":{"n":6,"median":0.281196167,"min":0.108888833,"max":0.438019625},"tokens_median":76.0,"slots":10},"prefill_call_ms":{"n":867,"median":25.238875,"min":24.255792000000003,"max":77.0695},"main_call_ms":{"n":850,"median":11.6534165,"min":11.100291,"max":22.617},"state_reset_ms":{"n":75,"median":0.49950000000000006,"min":0.43075,"max":2.095041},"footprint_mb":{"after_load":431.2384,"after_runs":839.249512,"max_during_runs":1011.412584}}},"loads":[{"pass":"int8mix_pf16_pass1","first_load_in_process":{"decoder_s":{"main":0.541543458,"model":0.001179083,"prefill":1.371166417},"tokenizer_s":0.8175804170000001,"tower_s":{"g256":0.171159959,"g448":0.9196118750000001},"wall_s":3.827721333},"reload_in_process":{"decoder_s":{"main":0.435518917,"model":0.0010166250000000002,"prefill":1.2044305},"tokenizer_s":0.63941375,"tower_s":{"g256":0.10903850000000001,"g448":0.112494833},"wall_s":2.505467416}},{"pass":"int8mix_pf16_pass2","first_load_in_process":{"decoder_s":{"main":0.429605292,"model":0.000959167,"prefill":1.086316625},"tokenizer_s":0.66205175,"tower_s":{"g256":0.14669879200000002,"g448":0.12058758300000001},"wall_s":2.450600333},"reload_in_process":{"decoder_s":{"main":0.495635541,"model":0.0010222080000000001,"prefill":1.6246965000000002},"tokenizer_s":0.724864833,"tower_s":{"g256":0.12984345800000002,"g448":0.13069587500000002},"wall_s":3.110846625}},{"pass":"fp16_pf16_pass1","first_load_in_process":{"decoder_s":{"main":4.454922292,"model":0.0019498750000000002,"prefill":12.933409458},"tokenizer_s":0.7293337080000001,"tower_s":{"g256":0.15357875,"g448":0.12195720800000001},"wall_s":18.399925208},"reload_in_process":{"decoder_s":{"main":0.5489210840000001,"model":0.000789458,"prefill":1.039204417},"tokenizer_s":0.627092375,"tower_s":{"g256":0.098451042,"g448":0.10132808300000001},"wall_s":2.418976125}},{"pass":"fp16_pf16_pass2","first_load_in_process":{"decoder_s":{"main":0.587681125,"model":0.0008149170000000001,"prefill":1.062625959},"tokenizer_s":0.6377271670000001,"tower_s":{"g256":0.128848041,"g448":0.099797333},"wall_s":2.521454458},"reload_in_process":{"decoder_s":{"main":0.563589416,"model":0.0008795420000000001,"prefill":1.07611625},"tokenizer_s":0.6718160830000001,"tower_s":{"g256":0.101566417,"g448":0.10481720900000001},"wall_s":2.522105583}}],"gpu_lock":{"lock":"/Users/majimadaisuke/code/coreai/_GPU_LOCK","started":"2026-09-29 03:06:51 JST","load_avg_start":[3.41357421875,4.54931640625,4.8720703125],"snapshot_before_lock":{"at":"before lock","time":"03:06:55","load_avg":[3.41357421875,4.54931640625,4.8720703125],"gpu_util_samples":[84,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:06:49","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"01:06:48","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":10.6,"rss_mb":78,"etime":"14-12:29:38","command":"WindowServer"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:41:28","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:34:59","command":"MTLCompilerService"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:29:25","command":"MTLCompilerService"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":0.4,"rss_mb":317,"etime":"03-14:01:30","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"29:25","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"25:03","command":"swift-plugin-server"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:19:49","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:26:32","command":"com.apple.dt.SKAgent"},{"pid":79647,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"11:30","command":"zsh"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:30:47","command":"taskgated"},{"pid":84652,"ppid":84651,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"06:46","command":"Python"},{"pid":84654,"ppid":84652,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"06:46","command":"Python"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:43:06","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"02:27","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"02:27","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:40:03","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:40:02","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:40:01","command":"SWBBuildService"},{"pid":92919,"ppid":92910,"pri":31,"cpu_pct":495.7,"rss_mb":7987,"etime":"00:17","command":"Python"},{"pid":93157,"ppid":84654,"pri":20,"cpu_pct":0.0,"rss_mb":63,"etime":"00:09","command":"Python"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:42:22","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":0.1,"rss_mb":425,"etime":"06:06:03","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:06:02","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":0.8,"rss_mb":540,"etime":"04:48:10","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:48:09","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":1.4,"rss_mb":552,"etime":"04:48:10","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:48:08","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":0.2,"rss_mb":384,"etime":"01:31:44","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:31:43","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:29:17","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":1.2,"rss_mb":585,"etime":"01:28:22","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:28:21","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":0.1,"rss_mb":568,"etime":"39:11","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"39:09","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"29:26","command":"sourcekit-lsp"}]},"file_existed":true,"file_bytes_before":0,"file_mtime_before":"2026-09-28 09:50:10","state":"taken by this run (flock; file 0 bytes, last written 2026-09-28 09:50:10; GPU utilization [0, 0, 0, 0, 0])","waited_s":4.1,"readings":[{"t_s":4.1,"flock_free":true,"tag":"","tag_pid_live":null,"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false}],"snapshot_after_lock":{"at":"after lock, before the passes","time":"03:07:04","load_avg":[3.82373046875,4.59716796875,4.884765625],"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:06:58","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"01:06:57","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":12.1,"rss_mb":78,"etime":"14-12:29:47","command":"WindowServer"},{"pid":671,"ppid":1,"pri":62,"cpu_pct":9.2,"rss_mb":69,"etime":"14-12:29:47","command":"loginwindow"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:41:37","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:35:08","command":"MTLCompilerService"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:29:34","command":"MTLCompilerService"},{"pid":30296,"ppid":1,"pri":4,"cpu_pct":6.6,"rss_mb":23,"etime":"07-04:51:40","command":"Storage"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":0.3,"rss_mb":317,"etime":"03-14:01:39","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"29:34","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"25:12","command":"swift-plugin-server"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:19:58","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:26:41","command":"com.apple.dt.SKAgent"},{"pid":79647,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"11:39","command":"zsh"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:30:56","command":"taskgated"},{"pid":84652,"ppid":84651,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"06:55","command":"Python"},{"pid":84654,"ppid":84652,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"06:55","command":"Python"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:43:15","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"02:36","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"02:36","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:40:12","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:40:11","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:40:10","command":"SWBBuildService"},{"pid":92919,"ppid":92910,"pri":31,"cpu_pct":461.2,"rss_mb":7998,"etime":"00:26","command":"Python"},{"pid":93157,"ppid":84654,"pri":20,"cpu_pct":0.0,"rss_mb":63,"etime":"00:18","command":"Python"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:42:31","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":0.0,"rss_mb":425,"etime":"06:06:12","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:06:11","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":1.0,"rss_mb":540,"etime":"04:48:19","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:48:18","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":0.9,"rss_mb":558,"etime":"04:48:19","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:48:17","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":0.1,"rss_mb":384,"etime":"01:31:53","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:31:52","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:29:26","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":2.1,"rss_mb":585,"etime":"01:28:31","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:28:30","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":5.6,"rss_mb":576,"etime":"39:20","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"39:18","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"29:35","command":"sourcekit-lsp"}]},"passes":[{"bundle":"decider_2b_vision_decode_int8mix_pf16","pass":1,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/int8mix_pf16_pass1.json","exit":0,"s":45.3,"priority_every_5s":[[0.0,31,3.82,0],[5.0,31,4.32,80],[10.1,31,5.17,84],[15.1,31,5.4,86],[20.1,31,5.29,86],[25.2,31,5.82,83],[30.2,31,6.64,85],[35.2,31,6.91,86],[40.3,31,6.43,85]],"priority_min":31,"demoted_samples":0},{"bundle":"decider_2b_vision_decode_int8mix_pf16","pass":2,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/int8mix_pf16_pass2.json","exit":0,"s":45.3,"priority_every_5s":[[0.0,31,6.08,83],[5.0,31,5.91,89],[10.1,31,5.6,89],[15.1,31,6.03,90],[20.1,31,5.95,89],[25.1,31,5.87,90],[30.2,31,5.8,90],[35.2,31,5.58,89],[40.2,31,5.85,80]],"priority_min":31,"demoted_samples":0},{"bundle":"decider_2b_vision_decode_fp16_pf16","pass":1,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/fp16_pf16_pass1.json","exit":0,"s":60.4,"priority_every_5s":[[0.0,31,5.78,0],[5.0,31,5.64,0],[10.1,46,5.35,0],[15.1,31,5.24,91],[20.1,31,5.38,92],[25.2,31,5.51,94],[30.2,31,6.19,91],[35.2,31,5.85,87],[40.2,31,5.62,87],[45.3,31,5.33,88],[50.3,31,5.31,89],[55.3,31,5.2,86]],"priority_min":31,"demoted_samples":0},{"bundle":"decider_2b_vision_decode_fp16_pf16","pass":2,"json":"/Users/majimadaisuke/code/coreai/_decider2bv/swift/timing/r8-d/fp16_pf16_pass2.json","exit":0,"s":45.3,"priority_every_5s":[[0.0,31,5.03,81],[5.1,31,5.02,87],[10.1,31,4.78,88],[15.1,31,4.48,87],[20.1,31,6.84,86],[25.2,31,6.69,87],[30.2,31,6.32,88],[35.2,31,6.29,87],[40.3,31,5.95,84]],"priority_min":31,"demoted_samples":0}],"snapshot_after_passes":{"at":"after the passes, lock still held","time":"03:10:24","load_avg":[5.5009765625,5.2734375,5.1318359375],"gpu_util_samples":[0,0,0,0,0],"gpu_busy":false,"processes":[{"pid":333,"ppid":1,"pri":20,"cpu_pct":0.0,"rss_mb":8,"etime":"01:10:18","command":"Python"},{"pid":345,"ppid":333,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"01:10:17","command":"Python"},{"pid":668,"ppid":1,"pri":79,"cpu_pct":18.1,"rss_mb":80,"etime":"14-12:33:07","command":"WindowServer"},{"pid":1258,"ppid":1,"pri":54,"cpu_pct":9.2,"rss_mb":21,"etime":"14-12:32:32","command":"WallpaperAerialsExtension"},{"pid":2528,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:44:57","command":"MTLCompilerService"},{"pid":11404,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":1,"etime":"01-15:38:28","command":"MTLCompilerService"},{"pid":14354,"ppid":1,"pri":46,"cpu_pct":15.5,"rss_mb":594,"etime":"05-02:59:28","command":"Google"},{"pid":17345,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"01-15:32:54","command":"MTLCompilerService"},{"pid":36522,"ppid":1,"pri":46,"cpu_pct":0.8,"rss_mb":318,"etime":"03-14:04:59","command":"Xcode"},{"pid":45968,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":268,"etime":"32:54","command":"SourceKitService"},{"pid":54672,"ppid":45968,"pri":31,"cpu_pct":0.0,"rss_mb":7,"etime":"28:32","command":"swift-plugin-server"},{"pid":59079,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":24,"etime":"01-13:23:18","command":"MTLCompilerService"},{"pid":59710,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":36,"etime":"03-09:30:01","command":"com.apple.dt.SKAgent"},{"pid":79647,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14:59","command":"zsh"},{"pid":83335,"ppid":1,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"14-07:34:16","command":"taskgated"},{"pid":84652,"ppid":84651,"pri":20,"cpu_pct":0.0,"rss_mb":20,"etime":"10:15","command":"Python"},{"pid":84654,"ppid":84652,"pri":20,"cpu_pct":0.0,"rss_mb":26,"etime":"10:15","command":"Python"},{"pid":88766,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":2,"etime":"13:46:35","command":"MTLCompilerService"},{"pid":90182,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":2,"etime":"05:56","command":"zsh"},{"pid":90184,"ppid":90182,"pri":31,"cpu_pct":0.0,"rss_mb":1,"etime":"05:56","command":"tail"},{"pid":90396,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":18,"etime":"13:43:32","command":"com.apple.dt.ExternalViewService"},{"pid":90407,"ppid":1,"pri":4,"cpu_pct":0.0,"rss_mb":4,"etime":"13:43:31","command":"com.apple.dt.Xcode.sourcecontrol.WorkingCopyScanner"},{"pid":90432,"ppid":36522,"pri":31,"cpu_pct":0.0,"rss_mb":16,"etime":"13:43:30","command":"SWBBuildService"},{"pid":93157,"ppid":84654,"pri":20,"cpu_pct":0.0,"rss_mb":63,"etime":"03:38","command":"Python"},{"pid":97574,"ppid":14354,"pri":4,"cpu_pct":77.7,"rss_mb":267,"etime":"00:01","command":"Google"},{"pid":40466,"ppid":40412,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"08:45:51","command":"python"},{"pid":86705,"ppid":86287,"pri":31,"cpu_pct":0.2,"rss_mb":426,"etime":"06:09:32","command":"claude"},{"pid":86811,"ppid":86705,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"06:09:31","command":"python"},{"pid":31193,"ppid":30900,"pri":31,"cpu_pct":0.1,"rss_mb":545,"etime":"04:51:39","command":"claude"},{"pid":31482,"ppid":31193,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:51:38","command":"python"},{"pid":31400,"ppid":31052,"pri":31,"cpu_pct":0.5,"rss_mb":587,"etime":"04:51:39","command":"claude"},{"pid":31511,"ppid":31400,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"04:51:37","command":"python"},{"pid":96617,"ppid":96397,"pri":31,"cpu_pct":7.6,"rss_mb":561,"etime":"00:33","command":"claude"},{"pid":96692,"ppid":96617,"pri":31,"cpu_pct":0.0,"rss_mb":61,"etime":"00:30","command":"python"},{"pid":56725,"ppid":56499,"pri":31,"cpu_pct":1.8,"rss_mb":386,"etime":"01:35:13","command":"claude"},{"pid":56779,"ppid":56725,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:35:12","command":"python"},{"pid":88333,"ppid":88266,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"02:32:46","command":"python"},{"pid":61818,"ppid":61343,"pri":31,"cpu_pct":2.0,"rss_mb":656,"etime":"01:31:51","command":"claude"},{"pid":62014,"ppid":61818,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"01:31:50","command":"python"},{"pid":31724,"ppid":31263,"pri":31,"cpu_pct":0.4,"rss_mb":585,"etime":"42:40","command":"claude"},{"pid":31797,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":12,"etime":"42:38","command":"python"},{"pid":45941,"ppid":31724,"pri":31,"cpu_pct":0.0,"rss_mb":109,"etime":"32:55","command":"sourcekit-lsp"}]},"load_avg_end":[5.5009765625,5.2734375,5.1318359375],"finished":"2026-09-29 03:10:24 JST","file_after":"0 bytes"},"contended":false,"definition":"wall_from_file = ImageIO decode of the PNG + RGB8 + resize + patches + tower + prompt + static inputs + decoder calls (chunk order) + read-out, per run (one row = one decision with 1..3 questions); the first run of each process is reported apart (first-call costs); load = AIModel(contentsOf:) + loadFunction, tokenizer apart","result":"done","note":"GPU lock taken (GPU 0% at the take), but another session's Python job (pid 92919, 496% CPU, 8 GB RSS, started 17 s before the take) shared the CPU during the passes (load average 5-7); fp16_pf16 is the alternative bundle, not shipped (round 7)"},"trimmed":{"full_copy":"~/code/coreai/_decider2bv/transcripts_full/gate-decider-2b-vision-swift.full.json","full_copy_sha256_at_trim":"e22585d56f07f203ef2c5c416d7b4769db0e1c1e5614265890f3a44410d80689","serialization":"compact json (no indentation) to keep the repository small; the full copy is pretty-printed"}}
