{ "schema": "coreai-letter-gate-summary/1", "result": "PASS", "recipe": "openjev-27b", "license": "CC BY-NC 4.0", "source": { "hf_id": "openjev/openjev", "revision": "5ec9e5fd2f80a6fff386779b1e5ac7e389971889", "helper_sha256": "81a22f1b1b8912a465059207ef9f60b7c6c16b4de6372305d867efbe38a1987a" }, "base": { "hf_id": "Qwen/Qwen3.8-27B", "revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0", "license": "Apache-2.0" }, "gate": "Author BF16 helper argmax at margin>=0.02; finite logits/probabilities; exact reset; both engines exact same-bundle full-vocabulary argmax text", "generated_at": "2026-09-23T05:30:53.145438+00:00", "bundle": "gpu-pipelined/openjev_27b_decode_int8hu_block32_sym", "protocol": { "layers": 64, "letters": "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz", "label_style": "bare", "template": "chat", "temperature": 0.85, "noul": { "t": 1.829074, "bias": 0 }, "slot": "len(ids)-1", "logits_shape": [ 1, 1, 248320 ], "max_evaluations_per_process": 6 }, "fixtures": { "path": "fixtures-openjev-27b.json", "sha256": "7a33b42c676f85bc4a99ef729b99788b2d7afd00ddb1ba370ea492f99c482863", "requests": 61, "rows": 61, "distinct_composed_prompts": 51 }, "fixture_coverage": { "requests": 61, "rows": 61, "ordinary_choice": 30, "large_menus_10_26": 6, "choice_27_52": 3, "long_descriptions": 6, "null_descriptions": 23, "noul": 14, "noul_explicit": 4, "score": 12, "score_10": 2, "multilingual": 4, "languages": [ "de", "fr", "ja", "zh" ], "agent_shape": 4, "zoo_only": 2, "state_token_range": [ 32, 108 ], "kit_prompt_token_range": [ 90, 505 ], "zoo_only_tokens": [ 1559, 1559 ] }, "scope": { "license": "CC BY-NC 4.0", "rows": 61, "distinct_composed_prompts": 51, "min_oracle_top2_margin": 0.9884852443474252, "scope": "Deterministic synthetic conversion parity coverage, not an accuracy or calibration benchmark. Repeated gate-state noul questions remain counted rows; raw evidence identifies every prompt." }, "transcripts": { "readout": "gate-openjev-27b-readout-int8hu.json", "engine": "gate-openjev-27b-engine-int8hu.json" }, "reference": { "name": "A: author helper / MLX BF16", "second_oracle": false, "helper_unchanged": true }, "accepted_round1": { "readout": { "rows": 61, "expected_rows": 61, "argmax_agreement": 61, "margin_ge_0_02_rows": 61, "margin_ge_0_02_argmax_agreement": 61, "near_ties": [], "high_margin_disagreements": [], "full_vocab_argmax_is_label": 61, "full_vocab_argmax_label_fraction": 1.0, "off_label_rows": [], "max_abs_delta_p": 0.00027640540650619716, "mean_of_row_mean_abs_delta_p": 2.0580679511791693e-05, "max_abs_delta_raw_logits": 0.1640625, "finite_all": true, "nonconstant_all": true, "reset_identical": true, "session_resets_identical": true, "provisional_int8hu_max_delta_p_expectation": 0.05, "provisional_expectation_exceeded": false, "provisional_expectation_policy": "Exceedance is recorded for supervisor judgment; no investigation or new tolerance.", "max_abs_delta_noul": 0.0003066579912641698, "max_process_peak_rss_bytes": 111894233088, "worst_five": [ { "id": "choice-null-3.decision", "primitive": "choice", "zoo_only": false, "tokens": 98, "slot": 97, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 22.046875, 15.8828125, 16.96875 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/readout_logits/int8hu-03-kit-a1/choice-null-3.decision.npy", "full_logits_sha256": "c4deb1ee3be7ae046aeea2ad5ce14c52c49292de0f2859d46f4acd1c9e2d00ae", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 32, "full_vocab_argmax_text": "A", "full_vocab_argmax_is_label": true, "option_argmax": 0, "letter_argmax": 0, "option_argmax_label": "A", "oracle_argmax": 0, "oracle_argmax_label": "A", "argmax_agrees": true, "p": [ 0.9967584013938904, 0.00070657191099599, 0.00253512361086905 ], "p_oracle": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "oracle_margin": 0.9937035879060697, "max_abs_delta_p": 0.00027640540650619716, "mean_abs_delta_p": 0.00018423796575228646, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.028645833333333332, "full_logits_min": -12.53125, "full_logits_max": 22.046875, "full_logits_absmax": 22.046875, "label_logits_min": 15.8828125, "label_logits_max": 22.046875, "label_logits_absmax": 22.046875, "oracle_logits_min": 15.875, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 121.03288445807993, "peak_rss_bytes": 91897462784, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "p_author": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "raw_logprobs": [ -0.009106529876589775, -6.173169136047363, -5.087231636047363 ], "author_raw_logprobs": [ -0.009790420532226562, -6.134790420532227, -5.009790420532227 ], "max_abs_delta_logprobs": 0.07744121551513672 }, { "id": "choice-null-1.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.203125, 15.1328125, 21.9375 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/readout_logits/int8hu-02-kit-a1/choice-null-1.decision.npy", "full_logits_sha256": "3b35506801d0ea97176255e98c0ef41fa7bb81615e142e513c72fc4bbcbc3100", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 34, "full_vocab_argmax_text": "C", "full_vocab_argmax_is_label": true, "option_argmax": 2, "letter_argmax": 2, "option_argmax_label": "C", "oracle_argmax": 2, "oracle_argmax_label": "C", "argmax_agrees": true, "p": [ 0.0011734227882698178, 0.00033311531296931207, 0.9984934329986572 ], "p_oracle": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "oracle_margin": 0.9977457167543653, "max_abs_delta_p": 0.0002470278566034656, "mean_abs_delta_p": 0.00016467560436775642, "max_abs_delta_raw_logits": 0.1328125, "mean_abs_delta_raw_logits": 0.09114583333333333, "full_logits_min": -13.140625, "full_logits_max": 21.9375, "full_logits_absmax": 21.9375, "label_logits_min": 15.1328125, "label_logits_max": 21.9375, "label_logits_absmax": 21.9375, "oracle_logits_min": 15.0, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 14.79682741605211, "peak_rss_bytes": 75378163712, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "p_author": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "raw_logprobs": [ -5.739787578582764, -6.810100078582764, -0.005412803962826729 ], "author_raw_logprobs": [ -5.879768371582031, -7.004768371582031, -0.00476837158203125 ], "max_abs_delta_logprobs": 0.19466829299926758 }, { "id": "score-01.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 13.1015625, 13.59375, 14.453125, 15.5625, 22.03125, 16.1875, 14.4140625, 13.4140625, 13.0234375, 12.453125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/readout_logits/int8hu-09-kit-a1/score-01.decision.npy", "full_logits_sha256": "536f747c4873b06680a00c639987755e5e3314eb04592d8cbfbd7005ebe2f104", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 36, "full_vocab_argmax_text": "E", "full_vocab_argmax_is_label": true, "option_argmax": 4, "letter_argmax": 4, "option_argmax_label": "E", "oracle_argmax": 4, "oracle_argmax_label": "E", "argmax_agrees": true, "p": [ 2.7331869205227122e-05, 4.8769092245493084e-05, 0.00013403836055658758, 0.0004943645908497274, 0.9980588555335999, 0.0010312917875126004, 0.00012801792763639241, 3.947630102629773e-05, 2.493172542017419e-05, 1.2745591448037885e-05 ], "p_oracle": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "oracle_margin": 0.9974363096831597, "max_abs_delta_p": 0.00023577690895315584, "mean_abs_delta_p": 4.713765974065759e-05, "max_abs_delta_raw_logits": 0.1015625, "mean_abs_delta_raw_logits": 0.05625, "full_logits_min": -13.5859375, "full_logits_max": 22.03125, "full_logits_absmax": 22.03125, "label_logits_min": 12.453125, "label_logits_max": 22.03125, "label_logits_absmax": 22.03125, "oracle_logits_min": 12.375, "oracle_logits_max": 22.125, "oracle_logits_absmax": 22.125, "finite": true, "nonconstant": true, "wall_seconds_contended": 32.56467883300502, "peak_rss_bytes": 91042463744, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "p_author": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "raw_logprobs": [ -8.936448097229004, -8.444260597229004, -7.584886074066162, -6.475511074066162, -0.006760936230421066, -5.850511074066162, -7.623948574066162, -8.623948097229004, -9.014573097229004, -9.584885597229004 ], "author_raw_logprobs": [ -9.006095886230469, -8.506095886230469, -7.631095886230469, -6.506095886230469, -0.00609588623046875, -6.006095886230469, -7.756095886230469, -8.818595886230469, -9.131095886230469, -9.756095886230469 ], "max_abs_delta_logprobs": 0.19464778900146484, "expected_index": 4.000550426457266, "author_expected_index": 4.000343465165037, "abs_delta_expected_index": 0.00020696129222930182 }, { "id": "score-00.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 15.3203125, 21.90625, 15.3828125, 14.5546875, 13.8515625, 13.7734375, 13.0546875, 12.953125, 12.3984375, 12.390625 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/readout_logits/int8hu-09-kit-a1/score-00.decision.npy", "full_logits_sha256": "30d6d7a959d95ccce8de11f43911727d7fe70c44cd4cc08d33f485537e116313", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0004309740907046944, 0.9986995458602905, 0.00046385693713091314, 0.00017509207827970386, 7.656242814846337e-05, 6.983917410252616e-05, 2.9982276828377508e-05, 2.6605568564264104e-05, 1.3853625205229037e-05, 1.3726886209042277e-05 ], "p_oracle": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "oracle_margin": 0.9985002872309481, "max_abs_delta_p": 0.0001832478789784986, "mean_abs_delta_p": 3.665346834209848e-05, "max_abs_delta_raw_logits": 0.09375, "mean_abs_delta_raw_logits": 0.04765625, "full_logits_min": -13.9453125, "full_logits_max": 21.90625, "full_logits_absmax": 21.90625, "label_logits_min": 12.390625, "label_logits_max": 21.90625, "label_logits_absmax": 21.90625, "oracle_logits_min": 12.3125, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 29.5946450419724, "peak_rss_bytes": 91042463744, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "p_author": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "raw_logprobs": [ -6.591176509857178, -0.005239089950919151, -6.528676509857178, -7.356801509857178, -8.059926986694336, -8.138051986694336, -8.856801986694336, -8.958364486694336, -9.513051986694336, -9.520864486694336 ], "author_raw_logprobs": [ -6.692146301269531, -0.00464630126953125, -6.692146301269531, -7.442146301269531, -8.192146301269531, -8.317146301269531, -9.004646301269531, -9.067146301269531, -9.629646301269531, -9.692146301269531 ], "max_abs_delta_logprobs": 0.1790943145751953, "expected_index": 1.0014084851709413, "author_expected_index": 1.0011797614695785, "abs_delta_expected_index": 0.0002287237013627319 }, { "id": "choice-null-2.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.015625, 21.546875, 17.03125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/readout_logits/int8hu-02-kit-a1/choice-null-2.decision.npy", "full_logits_sha256": "21dd3d64921dc91894588b1625d74bb99fe0d92335935380ca9b2a771dd5c236", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0014829020947217941, 0.9936190247535706, 0.00489813182502985 ], "p_oracle": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "oracle_margin": 0.9884852443474252, "max_abs_delta_p": 0.00014550740120000683, "mean_abs_delta_p": 9.698537635920142e-05, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.03125, "full_logits_min": -12.6015625, "full_logits_max": 21.546875, "full_logits_absmax": 21.546875, "label_logits_min": 16.015625, "label_logits_max": 21.546875, "label_logits_absmax": 21.546875, "oracle_logits_min": 16.0, "oracle_logits_max": 21.5, "oracle_logits_absmax": 21.5, "finite": true, "nonconstant": true, "wall_seconds_contended": 15.348673959029838, "peak_rss_bytes": 75378163712, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "p_author": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "raw_logprobs": [ -5.547797203063965, -0.016547108069062233, -4.532172203063965 ], "author_raw_logprobs": [ -5.517009735107422, -0.017009735107421875, -4.517009735107422 ], "max_abs_delta_logprobs": 0.03078746795654297 } ] }, "engine": { "engines": { "coreai-pipelined": { "rows": 61, "expected_rows": 61, "agreement": 61, "calls_succeeded": 61, "oracle_label_agreement": 61, "emits_row_label": 61, "disagreements": [], "oracle_label_disagreements": [], "first_call_preparation": { "id": "choice-menu-10.decision", "prepare_seconds_contended": 57.52, "model_load_seconds_contended": 57.5197, "cache_hit": false, "wall_seconds_contended": 69.50257504207548, "stdout_log": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/logs/engine-coreai-pipelined-int8hu-choice-menu-10.decision.log" } }, "coreai-sequential": { "rows": 61, "expected_rows": 61, "agreement": 61, "calls_succeeded": 61, "oracle_label_agreement": 61, "emits_row_label": 61, "disagreements": [], "oracle_label_disagreements": [], "first_call_preparation": { "id": "choice-menu-10.decision", "prepare_seconds_contended": 4.458, "model_load_seconds_contended": 4.458, "cache_hit": true, "wall_seconds_contended": 16.402605457929894, "stdout_log": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/logs/engine-coreai-sequential-int8hu-choice-menu-10.decision.log" } } }, "expected_calls": 122, "calls": 122, "agreement": 122, "oracle_label_agreement": 122 } }, "readout": { "rows": 61, "expected_rows": 61, "argmax_agreement": 61, "margin_ge_0_02_rows": 61, "margin_ge_0_02_argmax_agreement": 61, "near_ties": [], "high_margin_disagreements": [], "full_vocab_argmax_is_label": 61, "full_vocab_argmax_label_fraction": 1.0, "off_label_rows": [], "max_abs_delta_p": 0.00027640540650619716, "mean_of_row_mean_abs_delta_p": 2.0580679511791693e-05, "max_abs_delta_raw_logits": 0.1640625, "finite_all": true, "nonconstant_all": true, "reset_identical": true, "session_resets_identical": true, "provisional_int8hu_max_delta_p_expectation": 0.05, "provisional_max_delta_p_expectation": 0.05, "provisional_expectation_exceeded": false, "provisional_expectation_policy": "Exceedance is recorded for supervisor judgment; no investigation or new tolerance.", "max_abs_delta_noul": 0.0003066579912641698, "max_process_peak_rss_bytes": 116084408320, "worst_five": [ { "id": "choice-null-3.decision", "primitive": "choice", "zoo_only": false, "tokens": 98, "slot": 97, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 22.046875, 15.8828125, 16.96875 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-03-kit-a1/choice-null-3.decision.npy", "full_logits_sha256": "c4deb1ee3be7ae046aeea2ad5ce14c52c49292de0f2859d46f4acd1c9e2d00ae", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 32, "full_vocab_argmax_text": "A", "full_vocab_argmax_is_label": true, "option_argmax": 0, "letter_argmax": 0, "option_argmax_label": "A", "oracle_argmax": 0, "oracle_argmax_label": "A", "argmax_agrees": true, "p": [ 0.9967584013938904, 0.00070657191099599, 0.00253512361086905 ], "p_oracle": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "oracle_margin": 0.9937035879060697, "max_abs_delta_p": 0.00027640540650619716, "mean_abs_delta_p": 0.00018423796575228646, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.028645833333333332, "full_logits_min": -12.53125, "full_logits_max": 22.046875, "full_logits_absmax": 22.046875, "label_logits_min": 15.8828125, "label_logits_max": 22.046875, "label_logits_absmax": 22.046875, "oracle_logits_min": 15.875, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 140.64402829203755, "peak_rss_bytes": 81219371008, "author_argmax": 0, "author_argmax_label": "A", "author_argmax_agrees": true, "p_author": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "author_margin": 0.9937035879060697, "author_raw_logits": [ 22.0, 15.875, 17.0 ], "author_max_abs_delta_p": 0.00027640540650619716, "author_mean_abs_delta_p": 0.00018423796575228646, "author_max_abs_delta_raw_logits": 0.046875, "author_mean_abs_delta_raw_logits": 0.028645833333333332, "author_logits_min": 15.875, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 0.9999999999999999, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -0.009106529876589775, -6.173169136047363, -5.087231636047363 ], "author_raw_logprobs": [ -0.009790420532226562, -6.134790420532227, -5.009790420532227 ], "max_abs_delta_logprobs": 0.07744121551513672, "kind": "choice", "request_id": "choice-null-3" }, { "id": "choice-null-1.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.203125, 15.1328125, 21.9375 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-02-kit-a1/choice-null-1.decision.npy", "full_logits_sha256": "3b35506801d0ea97176255e98c0ef41fa7bb81615e142e513c72fc4bbcbc3100", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 34, "full_vocab_argmax_text": "C", "full_vocab_argmax_is_label": true, "option_argmax": 2, "letter_argmax": 2, "option_argmax_label": "C", "oracle_argmax": 2, "oracle_argmax_label": "C", "argmax_agrees": true, "p": [ 0.0011734227882698178, 0.00033311531296931207, 0.9984934329986572 ], "p_oracle": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "oracle_margin": 0.9977457167543653, "max_abs_delta_p": 0.0002470278566034656, "mean_abs_delta_p": 0.00016467560436775642, "max_abs_delta_raw_logits": 0.1328125, "mean_abs_delta_raw_logits": 0.09114583333333333, "full_logits_min": -13.140625, "full_logits_max": 21.9375, "full_logits_absmax": 21.9375, "label_logits_min": 15.1328125, "label_logits_max": 21.9375, "label_logits_absmax": 21.9375, "oracle_logits_min": 15.0, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 27.509680208051577, "peak_rss_bytes": 75650547712, "author_argmax": 2, "author_argmax_label": "C", "author_argmax_agrees": true, "p_author": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "author_margin": 0.9977457167543653, "author_raw_logits": [ 16.125, 15.0, 22.0 ], "author_max_abs_delta_p": 0.0002470278566034656, "author_mean_abs_delta_p": 0.00016467560436775642, "author_max_abs_delta_raw_logits": 0.1328125, "author_mean_abs_delta_raw_logits": 0.09114583333333333, "author_logits_min": 15.0, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 1.0, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -5.739787578582764, -6.810100078582764, -0.005412803962826729 ], "author_raw_logprobs": [ -5.879768371582031, -7.004768371582031, -0.00476837158203125 ], "max_abs_delta_logprobs": 0.19466829299926758, "kind": "choice", "request_id": "choice-null-1" }, { "id": "score-01.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 13.1015625, 13.59375, 14.453125, 15.5625, 22.03125, 16.1875, 14.4140625, 13.4140625, 13.0234375, 12.453125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-09-kit-a1/score-01.decision.npy", "full_logits_sha256": "536f747c4873b06680a00c639987755e5e3314eb04592d8cbfbd7005ebe2f104", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 36, "full_vocab_argmax_text": "E", "full_vocab_argmax_is_label": true, "option_argmax": 4, "letter_argmax": 4, "option_argmax_label": "E", "oracle_argmax": 4, "oracle_argmax_label": "E", "argmax_agrees": true, "p": [ 2.7331869205227122e-05, 4.8769092245493084e-05, 0.00013403836055658758, 0.0004943645908497274, 0.9980588555335999, 0.0010312917875126004, 0.00012801792763639241, 3.947630102629773e-05, 2.493172542017419e-05, 1.2745591448037885e-05 ], "p_oracle": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "oracle_margin": 0.9974363096831597, "max_abs_delta_p": 0.00023577690895315584, "mean_abs_delta_p": 4.713765974065759e-05, "max_abs_delta_raw_logits": 0.1015625, "mean_abs_delta_raw_logits": 0.05625, "full_logits_min": -13.5859375, "full_logits_max": 22.03125, "full_logits_absmax": 22.03125, "label_logits_min": 12.453125, "label_logits_max": 22.03125, "label_logits_absmax": 22.03125, "oracle_logits_min": 12.375, "oracle_logits_max": 22.125, "oracle_logits_absmax": 22.125, "finite": true, "nonconstant": true, "wall_seconds_contended": 23.120143292006105, "peak_rss_bytes": 109204013056, "author_argmax": 4, "author_argmax_label": "E", "author_argmax_agrees": true, "p_author": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "author_margin": 0.9974363096831597, "author_raw_logits": [ 13.125, 13.625, 14.5, 15.625, 22.125, 16.125, 14.375, 13.3125, 13.0, 12.375 ], "author_max_abs_delta_p": 0.00023577690895315584, "author_mean_abs_delta_p": 4.713765974065759e-05, "author_max_abs_delta_raw_logits": 0.1015625, "author_mean_abs_delta_raw_logits": 0.05625, "author_logits_min": 12.375, "author_logits_max": 22.125, "author_logits_absmax": 22.125, "author_probability_sum": 1.0, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -8.936448097229004, -8.444260597229004, -7.584886074066162, -6.475511074066162, -0.006760936230421066, -5.850511074066162, -7.623948574066162, -8.623948097229004, -9.014573097229004, -9.584885597229004 ], "author_raw_logprobs": [ -9.006095886230469, -8.506095886230469, -7.631095886230469, -6.506095886230469, -0.00609588623046875, -6.006095886230469, -7.756095886230469, -8.818595886230469, -9.131095886230469, -9.756095886230469 ], "max_abs_delta_logprobs": 0.19464778900146484, "expected_index": 4.000550426457266, "author_expected_index": 4.000343465165037, "abs_delta_expected_index": 0.00020696129222930182, "kind": "score", "request_id": "score-01" }, { "id": "score-00.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 15.3203125, 21.90625, 15.3828125, 14.5546875, 13.8515625, 13.7734375, 13.0546875, 12.953125, 12.3984375, 12.390625 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-09-kit-a1/score-00.decision.npy", "full_logits_sha256": "30d6d7a959d95ccce8de11f43911727d7fe70c44cd4cc08d33f485537e116313", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0004309740907046944, 0.9986995458602905, 0.00046385693713091314, 0.00017509207827970386, 7.656242814846337e-05, 6.983917410252616e-05, 2.9982276828377508e-05, 2.6605568564264104e-05, 1.3853625205229037e-05, 1.3726886209042277e-05 ], "p_oracle": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "oracle_margin": 0.9985002872309481, "max_abs_delta_p": 0.0001832478789784986, "mean_abs_delta_p": 3.665346834209848e-05, "max_abs_delta_raw_logits": 0.09375, "mean_abs_delta_raw_logits": 0.04765625, "full_logits_min": -13.9453125, "full_logits_max": 21.90625, "full_logits_absmax": 21.90625, "label_logits_min": 12.390625, "label_logits_max": 21.90625, "label_logits_absmax": 21.90625, "oracle_logits_min": 12.3125, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 22.98662258393597, "peak_rss_bytes": 109204013056, "author_argmax": 1, "author_argmax_label": "B", "author_argmax_agrees": true, "p_author": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "author_margin": 0.9985002872309481, "author_raw_logits": [ 15.3125, 22.0, 15.3125, 14.5625, 13.8125, 13.6875, 13.0, 12.9375, 12.375, 12.3125 ], "author_max_abs_delta_p": 0.0001832478789784986, "author_mean_abs_delta_p": 3.665346834209848e-05, "author_max_abs_delta_raw_logits": 0.09375, "author_mean_abs_delta_raw_logits": 0.04765625, "author_logits_min": 12.3125, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 0.9999999999999998, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -6.591176509857178, -0.005239089950919151, -6.528676509857178, -7.356801509857178, -8.059926986694336, -8.138051986694336, -8.856801986694336, -8.958364486694336, -9.513051986694336, -9.520864486694336 ], "author_raw_logprobs": [ -6.692146301269531, -0.00464630126953125, -6.692146301269531, -7.442146301269531, -8.192146301269531, -8.317146301269531, -9.004646301269531, -9.067146301269531, -9.629646301269531, -9.692146301269531 ], "max_abs_delta_logprobs": 0.1790943145751953, "expected_index": 1.0014084851709413, "author_expected_index": 1.0011797614695785, "abs_delta_expected_index": 0.0002287237013627319, "kind": "score", "request_id": "score-00" }, { "id": "choice-null-2.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.015625, 21.546875, 17.03125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-02-kit-a1/choice-null-2.decision.npy", "full_logits_sha256": "21dd3d64921dc91894588b1625d74bb99fe0d92335935380ca9b2a771dd5c236", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0014829020947217941, 0.9936190247535706, 0.00489813182502985 ], "p_oracle": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "oracle_margin": 0.9884852443474252, "max_abs_delta_p": 0.00014550740120000683, "mean_abs_delta_p": 9.698537635920142e-05, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.03125, "full_logits_min": -12.6015625, "full_logits_max": 21.546875, "full_logits_absmax": 21.546875, "label_logits_min": 16.015625, "label_logits_max": 21.546875, "label_logits_absmax": 21.546875, "oracle_logits_min": 16.0, "oracle_logits_max": 21.5, "oracle_logits_absmax": 21.5, "finite": true, "nonconstant": true, "wall_seconds_contended": 27.36539225000888, "peak_rss_bytes": 75650547712, "author_argmax": 1, "author_argmax_label": "B", "author_argmax_agrees": true, "p_author": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "author_margin": 0.9884852443474252, "author_raw_logits": [ 16.0, 21.5, 17.0 ], "author_max_abs_delta_p": 0.00014550740120000683, "author_mean_abs_delta_p": 9.698537635920142e-05, "author_max_abs_delta_raw_logits": 0.046875, "author_mean_abs_delta_raw_logits": 0.03125, "author_logits_min": 16.0, "author_logits_max": 21.5, "author_logits_absmax": 21.5, "author_probability_sum": 0.9999999999999998, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -5.547797203063965, -0.016547108069062233, -4.532172203063965 ], "author_raw_logprobs": [ -5.517009735107422, -0.017009735107421875, -4.517009735107422 ], "max_abs_delta_logprobs": 0.03078746795654297, "kind": "choice", "request_id": "choice-null-2" } ], "author_rows": 61, "author_argmax_agreement": 61, "author_margin_ge_0_02_rows": 61, "author_margin_ge_0_02_argmax_agreement": 61, "author_max_abs_delta_p": 0.00027640540650619716, "author_mean_of_row_mean_abs_delta_p": 2.0580679511791693e-05, "author_max_abs_delta_raw_logits": 0.1640625, "author_near_ties": [], "author_disagreements": [], "author_worst_five": [ { "id": "choice-null-3.decision", "primitive": "choice", "zoo_only": false, "tokens": 98, "slot": 97, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 22.046875, 15.8828125, 16.96875 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-03-kit-a1/choice-null-3.decision.npy", "full_logits_sha256": "c4deb1ee3be7ae046aeea2ad5ce14c52c49292de0f2859d46f4acd1c9e2d00ae", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 32, "full_vocab_argmax_text": "A", "full_vocab_argmax_is_label": true, "option_argmax": 0, "letter_argmax": 0, "option_argmax_label": "A", "oracle_argmax": 0, "oracle_argmax_label": "A", "argmax_agrees": true, "p": [ 0.9967584013938904, 0.00070657191099599, 0.00253512361086905 ], "p_oracle": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "oracle_margin": 0.9937035879060697, "max_abs_delta_p": 0.00027640540650619716, "mean_abs_delta_p": 0.00018423796575228646, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.028645833333333332, "full_logits_min": -12.53125, "full_logits_max": 22.046875, "full_logits_absmax": 22.046875, "label_logits_min": 15.8828125, "label_logits_max": 22.046875, "label_logits_absmax": 22.046875, "oracle_logits_min": 15.875, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 140.64402829203755, "peak_rss_bytes": 81219371008, "author_argmax": 0, "author_argmax_label": "A", "author_argmax_agrees": true, "p_author": [ 0.9964819959873842, 0.0007395959313012184, 0.002778408081314484 ], "author_margin": 0.9937035879060697, "author_raw_logits": [ 22.0, 15.875, 17.0 ], "author_max_abs_delta_p": 0.00027640540650619716, "author_mean_abs_delta_p": 0.00018423796575228646, "author_max_abs_delta_raw_logits": 0.046875, "author_mean_abs_delta_raw_logits": 0.028645833333333332, "author_logits_min": 15.875, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 0.9999999999999999, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -0.009106529876589775, -6.173169136047363, -5.087231636047363 ], "author_raw_logprobs": [ -0.009790420532226562, -6.134790420532227, -5.009790420532227 ], "max_abs_delta_logprobs": 0.07744121551513672, "kind": "choice", "request_id": "choice-null-3" }, { "id": "choice-null-1.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.203125, 15.1328125, 21.9375 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-02-kit-a1/choice-null-1.decision.npy", "full_logits_sha256": "3b35506801d0ea97176255e98c0ef41fa7bb81615e142e513c72fc4bbcbc3100", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 34, "full_vocab_argmax_text": "C", "full_vocab_argmax_is_label": true, "option_argmax": 2, "letter_argmax": 2, "option_argmax_label": "C", "oracle_argmax": 2, "oracle_argmax_label": "C", "argmax_agrees": true, "p": [ 0.0011734227882698178, 0.00033311531296931207, 0.9984934329986572 ], "p_oracle": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "oracle_margin": 0.9977457167543653, "max_abs_delta_p": 0.0002470278566034656, "mean_abs_delta_p": 0.00016467560436775642, "max_abs_delta_raw_logits": 0.1328125, "mean_abs_delta_raw_logits": 0.09114583333333333, "full_logits_min": -13.140625, "full_logits_max": 21.9375, "full_logits_absmax": 21.9375, "label_logits_min": 15.1328125, "label_logits_max": 21.9375, "label_logits_absmax": 21.9375, "oracle_logits_min": 15.0, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 27.509680208051577, "peak_rss_bytes": 75650547712, "author_argmax": 2, "author_argmax_label": "C", "author_argmax_agrees": true, "p_author": [ 0.000994744100895361, 0.00026479504384396537, 0.9987404608552607 ], "author_margin": 0.9977457167543653, "author_raw_logits": [ 16.125, 15.0, 22.0 ], "author_max_abs_delta_p": 0.0002470278566034656, "author_mean_abs_delta_p": 0.00016467560436775642, "author_max_abs_delta_raw_logits": 0.1328125, "author_mean_abs_delta_raw_logits": 0.09114583333333333, "author_logits_min": 15.0, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 1.0, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -5.739787578582764, -6.810100078582764, -0.005412803962826729 ], "author_raw_logprobs": [ -5.879768371582031, -7.004768371582031, -0.00476837158203125 ], "max_abs_delta_logprobs": 0.19466829299926758, "kind": "choice", "request_id": "choice-null-1" }, { "id": "score-01.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 13.1015625, 13.59375, 14.453125, 15.5625, 22.03125, 16.1875, 14.4140625, 13.4140625, 13.0234375, 12.453125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-09-kit-a1/score-01.decision.npy", "full_logits_sha256": "536f747c4873b06680a00c639987755e5e3314eb04592d8cbfbd7005ebe2f104", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 36, "full_vocab_argmax_text": "E", "full_vocab_argmax_is_label": true, "option_argmax": 4, "letter_argmax": 4, "option_argmax_label": "E", "oracle_argmax": 4, "oracle_argmax_label": "E", "argmax_agrees": true, "p": [ 2.7331869205227122e-05, 4.8769092245493084e-05, 0.00013403836055658758, 0.0004943645908497274, 0.9980588555335999, 0.0010312917875126004, 0.00012801792763639241, 3.947630102629773e-05, 2.493172542017419e-05, 1.2745591448037885e-05 ], "p_oracle": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "oracle_margin": 0.9974363096831597, "max_abs_delta_p": 0.00023577690895315584, "mean_abs_delta_p": 4.713765974065759e-05, "max_abs_delta_raw_logits": 0.1015625, "mean_abs_delta_raw_logits": 0.05625, "full_logits_min": -13.5859375, "full_logits_max": 22.03125, "full_logits_absmax": 22.03125, "label_logits_min": 12.453125, "label_logits_max": 22.03125, "label_logits_absmax": 22.03125, "oracle_logits_min": 12.375, "oracle_logits_max": 22.125, "oracle_logits_absmax": 22.125, "finite": true, "nonconstant": true, "wall_seconds_contended": 23.120143292006105, "peak_rss_bytes": 109204013056, "author_argmax": 4, "author_argmax_label": "E", "author_argmax_agrees": true, "p_author": [ 2.5167876530591373e-05, 4.532250619515754e-05, 0.00012687666835658028, 0.0004766320983837456, 0.998294632442553, 0.0008583227593933479, 0.00010952542080557736, 3.137955843892878e-05, 2.1725998195734203e-05, 1.0414671147454758e-05 ], "author_margin": 0.9974363096831597, "author_raw_logits": [ 13.125, 13.625, 14.5, 15.625, 22.125, 16.125, 14.375, 13.3125, 13.0, 12.375 ], "author_max_abs_delta_p": 0.00023577690895315584, "author_mean_abs_delta_p": 4.713765974065759e-05, "author_max_abs_delta_raw_logits": 0.1015625, "author_mean_abs_delta_raw_logits": 0.05625, "author_logits_min": 12.375, "author_logits_max": 22.125, "author_logits_absmax": 22.125, "author_probability_sum": 1.0, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -8.936448097229004, -8.444260597229004, -7.584886074066162, -6.475511074066162, -0.006760936230421066, -5.850511074066162, -7.623948574066162, -8.623948097229004, -9.014573097229004, -9.584885597229004 ], "author_raw_logprobs": [ -9.006095886230469, -8.506095886230469, -7.631095886230469, -6.506095886230469, -0.00609588623046875, -6.006095886230469, -7.756095886230469, -8.818595886230469, -9.131095886230469, -9.756095886230469 ], "max_abs_delta_logprobs": 0.19464778900146484, "expected_index": 4.000550426457266, "author_expected_index": 4.000343465165037, "abs_delta_expected_index": 0.00020696129222930182, "kind": "score", "request_id": "score-01" }, { "id": "score-00.decision", "primitive": "score", "zoo_only": false, "tokens": 186, "slot": 185, "nopts": 10, "labels": [ "A", "B", "C", "D", "E", "F", "G", "H", "I", "J" ], "label_ids": [ 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 ], "temperature": 0.85, "raw_logits": [ 15.3203125, 21.90625, 15.3828125, 14.5546875, 13.8515625, 13.7734375, 13.0546875, 12.953125, 12.3984375, 12.390625 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-09-kit-a1/score-00.decision.npy", "full_logits_sha256": "30d6d7a959d95ccce8de11f43911727d7fe70c44cd4cc08d33f485537e116313", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0004309740907046944, 0.9986995458602905, 0.00046385693713091314, 0.00017509207827970386, 7.656242814846337e-05, 6.983917410252616e-05, 2.9982276828377508e-05, 2.6605568564264104e-05, 1.3853625205229037e-05, 1.3726886209042277e-05 ], "p_oracle": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "oracle_margin": 0.9985002872309481, "max_abs_delta_p": 0.0001832478789784986, "mean_abs_delta_p": 3.665346834209848e-05, "max_abs_delta_raw_logits": 0.09375, "mean_abs_delta_raw_logits": 0.04765625, "full_logits_min": -13.9453125, "full_logits_max": 21.90625, "full_logits_absmax": 21.90625, "label_logits_min": 12.390625, "label_logits_max": 21.90625, "label_logits_absmax": 21.90625, "oracle_logits_min": 12.3125, "oracle_logits_max": 22.0, "oracle_logits_absmax": 22.0, "finite": true, "nonconstant": true, "wall_seconds_contended": 22.98662258393597, "peak_rss_bytes": 109204013056, "author_argmax": 1, "author_argmax_label": "B", "author_argmax_agrees": true, "p_author": [ 0.0003825065083209087, 0.998882793739269, 0.0003825065083209087, 0.0001582842911312479, 6.549932164265925e-05, 5.6541843810324497e-05, 2.5182704588776518e-05, 2.3397472911134162e-05, 1.2071693301857253e-05, 1.121591670291069e-05 ], "author_margin": 0.9985002872309481, "author_raw_logits": [ 15.3125, 22.0, 15.3125, 14.5625, 13.8125, 13.6875, 13.0, 12.9375, 12.375, 12.3125 ], "author_max_abs_delta_p": 0.0001832478789784986, "author_mean_abs_delta_p": 3.665346834209848e-05, "author_max_abs_delta_raw_logits": 0.09375, "author_mean_abs_delta_raw_logits": 0.04765625, "author_logits_min": 12.3125, "author_logits_max": 22.0, "author_logits_absmax": 22.0, "author_probability_sum": 0.9999999999999998, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -6.591176509857178, -0.005239089950919151, -6.528676509857178, -7.356801509857178, -8.059926986694336, -8.138051986694336, -8.856801986694336, -8.958364486694336, -9.513051986694336, -9.520864486694336 ], "author_raw_logprobs": [ -6.692146301269531, -0.00464630126953125, -6.692146301269531, -7.442146301269531, -8.192146301269531, -8.317146301269531, -9.004646301269531, -9.067146301269531, -9.629646301269531, -9.692146301269531 ], "max_abs_delta_logprobs": 0.1790943145751953, "expected_index": 1.0014084851709413, "author_expected_index": 1.0011797614695785, "abs_delta_expected_index": 0.0002287237013627319, "kind": "score", "request_id": "score-00" }, { "id": "choice-null-2.decision", "primitive": "choice", "zoo_only": false, "tokens": 92, "slot": 91, "nopts": 3, "labels": [ "A", "B", "C" ], "label_ids": [ 32, 33, 34 ], "temperature": 0.85, "raw_logits": [ 16.015625, 21.546875, 17.03125 ], "full_logits_npy": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/results/readout_logits/int8hu-02-kit-a1/choice-null-2.decision.npy", "full_logits_sha256": "21dd3d64921dc91894588b1625d74bb99fe0d92335935380ca9b2a771dd5c236", "full_logits_dtype": "float16", "full_logits_shape": [ 248320 ], "full_vocab_argmax_id": 33, "full_vocab_argmax_text": "B", "full_vocab_argmax_is_label": true, "option_argmax": 1, "letter_argmax": 1, "option_argmax_label": "B", "oracle_argmax": 1, "oracle_argmax_label": "B", "argmax_agrees": true, "p": [ 0.0014829020947217941, 0.9936190247535706, 0.00489813182502985 ], "p_oracle": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "oracle_margin": 0.9884852443474252, "max_abs_delta_p": 0.00014550740120000683, "mean_abs_delta_p": 9.698537635920142e-05, "max_abs_delta_raw_logits": 0.046875, "mean_abs_delta_raw_logits": 0.03125, "full_logits_min": -12.6015625, "full_logits_max": 21.546875, "full_logits_absmax": 21.546875, "label_logits_min": 16.015625, "label_logits_max": 21.546875, "label_logits_absmax": 21.546875, "oracle_logits_min": 16.0, "oracle_logits_max": 21.5, "oracle_logits_absmax": 21.5, "finite": true, "nonconstant": true, "wall_seconds_contended": 27.36539225000888, "peak_rss_bytes": 75650547712, "author_argmax": 1, "author_argmax_label": "B", "author_argmax_agrees": true, "p_author": [ 0.0015382096426839417, 0.9934735173523705, 0.0049882730049453 ], "author_margin": 0.9884852443474252, "author_raw_logits": [ 16.0, 21.5, 17.0 ], "author_max_abs_delta_p": 0.00014550740120000683, "author_mean_abs_delta_p": 9.698537635920142e-05, "author_max_abs_delta_raw_logits": 0.046875, "author_mean_abs_delta_raw_logits": 0.03125, "author_logits_min": 16.0, "author_logits_max": 21.5, "author_logits_absmax": 21.5, "author_probability_sum": 0.9999999999999998, "license": "CC BY-NC 4.0", "oracle": "A: author helper / MLX bf16", "raw_logprobs": [ -5.547797203063965, -0.016547108069062233, -4.532172203063965 ], "author_raw_logprobs": [ -5.517009735107422, -0.017009735107421875, -4.517009735107422 ], "max_abs_delta_logprobs": 0.03078746795654297, "kind": "choice", "request_id": "choice-null-2" } ] }, "engine": { "engines": { "coreai-pipelined": { "rows": 61, "expected_rows": 61, "agreement": 61, "calls_succeeded": 61, "oracle_label_agreement": 61, "emits_row_label": 61, "disagreements": [], "oracle_label_disagreements": [], "first_call_preparation": { "id": "choice-menu-10.decision", "prepare_seconds_contended": 41.617, "model_load_seconds_contended": 41.617, "cache_hit": true, "wall_seconds_contended": 53.25791675003711, "stdout_log": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/logs/engine-coreai-pipelined-int8hu-choice-menu-10.decision.log" }, "author_rows": 61, "author_label_agreement": 61 }, "coreai-sequential": { "rows": 61, "expected_rows": 61, "agreement": 61, "calls_succeeded": 61, "oracle_label_agreement": 61, "emits_row_label": 61, "disagreements": [], "oracle_label_disagreements": [], "first_call_preparation": { "id": "choice-menu-10.decision", "prepare_seconds_contended": 4.818, "model_load_seconds_contended": 4.8178, "cache_hit": true, "wall_seconds_contended": 16.961319916998036, "stdout_log": "/Users/majimadaisuke/code/codex-conversions/2026-09-23/openjev-27b-coreai/results/round2/gate-work/logs/engine-coreai-sequential-int8hu-choice-menu-10.decision.log" }, "author_rows": 61, "author_label_agreement": 61 } }, "expected_calls": 122, "calls": 122, "agreement": 122, "oracle_label_agreement": 122, "author_rows": 122, "author_label_agreement": 122 }, "round2_reproduction": { "status": "PASS", "required_max_delta_p_difference_at_most": 1e-06, "option_argmax": 61, "expected_rows": 61, "max_abs_delta_p_difference": 0.0, "engine_agreement": 122 }, "artifact": { "name": "openjev_27b_decode_int8hu_block32_sym", "bundle": "gpu-pipelined/openjev_27b_decode_int8hu_block32_sym", "bytes": 29803295665, "main_bytes": 29774700211, "main_sha256": "004d52234efd0ac6ef2f7d81c33d72a502ad8a768d5424448049fcf6d3b73cc5", "metadata_bytes": 887, "metadata_sha256": "6f15f99cc3e25505d63fb5b086dcd189d9b9684ac13fab6468b513499aa1be1c" }, "export": { "successful_exports": 1, "round2_reexport": false, "reason": "Verified source weight blobs deleted after round1 gates; exporter diff and metadata contract are reproduction evidence", "wall_seconds": 234.33582841604948, "peak_rss_bytes": 72858877952, "round3_reexport": false, "metadata_flag_validation": { "result": "PASS", "equal_except": "compilation.date", "flags": [ "--revision", "--extra-metadata" ], "graph_unchanged": true } }, "readout_peak_rss_bytes_round1": 111894233088, "readout_peak_rss_bytes_round2": 116084408320, "kit_status": "measured through coreai-kit; supervisor supplied; all kit times contended", "measurements_coreai_kit": { "path": "measurements-coreai-kit.json", "sha256": "b109df4d9e654c7c2bc31cd289ed67d780c0ddf454917697822dd693cb32c8c1", "measured_by_this_run": false }, "throughput_proxy": { "status": "NOT MEASURED", "reason": "No quiet GPU window existed on 2026-09-23; the supervisor withdrew this requirement in round 3." } }