{ "benchmark": "RuBench v1", "split": "dev", "n_answers": 200, "n_criteria": 604, "built_at": "2026-11-04T00:15:10", "conditions": "first turn (feedback_mode=hint_1, reveal_requested=false, history=[]), temperature 0, max_tokens 1024, enable_thinking=false, no JSON schema enforcement. RuBio: SYSTEM_PROMPT only; base: SYSTEM_PROMPT+POLICY. Invalid JSON counts as wrong.", "finetuned": { "key": "8b-v6-s3407", "label": "RuBio-8B", "hf": "mulkzz/RuBio-8B@21e6bf3d", "in_dist_train_dev_criterion_match": 0.9395109395109396, "name": "RuBio-8B RuIso-v6 (시드 3407)", "answers": 200, "missing_output": 0, "json_invalid": 1, "criterion_acc": { "mean": 0.7595833333333333, "ci95": [ 0.7214769175142441, 0.7976897491524224 ], "n_answers": 200 }, "criterion_acc_excl_r2_2of3": { "mean": 0.7816666666666667, "ci95": [ 0.7422055735860975, 0.821127759747236 ], "n_answers": 200 }, "criterion_micro": { "hit": 458, "n": 604 }, "misconception": { "k": 86, "n": 133, "p": 0.6466165413533834, "ci95": [ 0.5623037851806594, 0.7226975936965256 ] }, "misconception_with_id": { "k": 53, "n": 79, "p": 0.6708860759493671, "ci95": [ 0.5614659956181937, 0.7644578289377567 ] }, "json_valid": { "k": 199, "n": 200, "p": 0.995, "ci95": [ 0.9722263725443985, 0.9991168287700527 ] }, "unit_criterion_micro": { "I": { "k": 153, "n": 206, "p": 0.7427184466019418, "ci95": [ 0.6789634796217361, 0.7975868074630104 ] }, "II": { "k": 154, "n": 200, "p": 0.77, "ci95": [ 0.7069170932155301, 0.822906471137807 ] }, "III": { "k": 151, "n": 198, "p": 0.7626262626262627, "ci95": [ 0.6987187578979108, 0.8165371694793246 ] } }, "other_issues": { "k": 155, "n": 200, "p": 0.775, "ci95": [ 0.7122589040583975, 0.8273762077829643 ] }, "answer_status": { "k": 125, "n": 200, "p": 0.625, "ci95": [ 0.5561431792007809, 0.6891455079998381 ] }, "gold_tier_criterion_acc": { "r1-3/3": { "k": 422, "n": 521, "p": 0.8099808061420346, "ci95": [ 0.7740717294391473, 0.8413522318045736 ] }, "r2-3/3": { "k": 13, "n": 23, "p": 0.5652173913043478, "ci95": [ 0.36811411127539473, 0.7436533513557193 ] }, "r2-2/3": { "k": 19, "n": 51, "p": 0.37254901960784315, "ci95": [ 0.25320540458646623, 0.5097475878756154 ] }, "human": { "k": 4, "n": 9, "p": 0.4444444444444444, "ci95": [ 0.18877886314051356, 0.7333482697447213 ] } }, "other_issues_split": { "has_error_recall": { "k": 35, "n": 53, "p": 0.660377358490566, "ci95": [ 0.5259481435437473, 0.7731294140000674 ] }, "no_error_specificity": { "k": 120, "n": 147, "p": 0.8163265306122449, "ci95": [ 0.745954432211289, 0.8705870037898586 ] } }, "confusion": { "contradicted->(없음)": 1, "contradicted->contradicted": 107, "contradicted->missing": 13, "contradicted->supported": 8, "contradicted->unclear": 4, "missing->(없음)": 1, "missing->contradicted": 21, "missing->missing": 98, "missing->supported": 18, "missing->unclear": 4, "supported->(없음)": 2, "supported->contradicted": 16, "supported->missing": 18, "supported->supported": 241, "supported->unclear": 6, "unclear->(없음)": 1, "unclear->contradicted": 9, "unclear->missing": 11, "unclear->supported": 13, "unclear->unclear": 12 } }, "base": { "key": "8b-base", "label": "Qwen3-8B-MegaScience (unfenced)", "hf": "MegaScience/Qwen3-8B-MegaScience", "serving": "RunPod A40 Pod, vLLM 0.32.0, bf16", "name": "Qwen3-8B-MegaScience (기본)", "answers": 200, "missing_output": 0, "json_invalid": 8, "criterion_acc": { "mean": 0.5620833333333333, "ci95": [ 0.5168683002478879, 0.6072983664187787 ], "n_answers": 200 }, "criterion_acc_excl_r2_2of3": { "mean": 0.5779166666666666, "ci95": [ 0.5315152183706893, 0.624318114962644 ], "n_answers": 200 }, "criterion_micro": { "hit": 342, "n": 604 }, "misconception": { "k": 26, "n": 133, "p": 0.19548872180451127, "ci95": [ 0.1370439507829323, 0.2710301084721449 ] }, "misconception_with_id": { "k": 26, "n": 79, "p": 0.3291139240506329, "ci95": [ 0.23554217106224323, 0.4385340043818064 ] }, "json_valid": { "k": 192, "n": 200, "p": 0.96, "ci95": [ 0.9230680280606571, 0.9795943408376206 ] }, "unit_criterion_micro": { "I": { "k": 120, "n": 206, "p": 0.5825242718446602, "ci95": [ 0.5142736573106079, 0.6477534402982059 ] }, "II": { "k": 111, "n": 200, "p": 0.555, "ci95": [ 0.4857331643905435, 0.622193857977729 ] }, "III": { "k": 111, "n": 198, "p": 0.5606060606060606, "ci95": [ 0.49097326500716265, 0.6279319490029686 ] } }, "other_issues": { "k": 141, "n": 200, "p": 0.705, "ci95": [ 0.6384129957526774, 0.7638604512563376 ] }, "answer_status": { "k": 44, "n": 200, "p": 0.22, "ci95": [ 0.16816547742776328, 0.2823878632428503 ] }, "gold_tier_criterion_acc": { "r1-3/3": { "k": 318, "n": 521, "p": 0.6103646833013435, "ci95": [ 0.5678278575107883, 0.651285936740072 ] }, "r2-3/3": { "k": 4, "n": 23, "p": 0.17391304347826086, "ci95": [ 0.06978666088359675, 0.37137602596083263 ] }, "r2-2/3": { "k": 17, "n": 51, "p": 0.3333333333333333, "ci95": [ 0.21969951335330842, 0.4703159383278752 ] }, "human": { "k": 3, "n": 9, "p": 0.3333333333333333, "ci95": [ 0.12058406396498972, 0.6457973346907151 ] } }, "other_issues_split": { "has_error_recall": { "k": 0, "n": 53, "p": 0.0, "ci95": [ 0.0, 0.06758173236340116 ] }, "no_error_specificity": { "k": 141, "n": 147, "p": 0.9591836734693877, "ci95": [ 0.9138180937286629, 0.9811614101439707 ] } }, "confusion": { "contradicted->(없음)": 2, "contradicted->contradicted": 37, "contradicted->missing": 21, "contradicted->supported": 73, "missing->(없음)": 7, "missing->contradicted": 7, "missing->missing": 42, "missing->supported": 86, "supported->(없음)": 13, "supported->contradicted": 2, "supported->missing": 5, "supported->supported": 263, "unclear->contradicted": 3, "unclear->missing": 8, "unclear->supported": 35 }, "parse": "unfenced (```json code fence removed before json.loads)", "strict": { "criterion_acc": { "mean": 0.0, "ci95": [ 0.0, 0.0 ], "n_answers": 200 }, "json_invalid": 200, "note": "official strict scoring (fenced JSON counts as invalid)" } }, "paired_ft_minus_base": { "all": { "diff": 0.1975, "ci95": [ 0.1481006807023753, 0.2468993192976247 ], "n": 200, "verdict": "개선" }, "no_majority": { "diff": 0.20375, "ci95": [ 0.1486619394502655, 0.25883806054973446 ], "n": 200, "verdict": "개선" } } }