{
 "run_id": "run-20260910T064154-bada0b",
 "started_at": 1789022514,
 "official": true,
 "page": "https://ranking.ai-roundtable.de/runs/run-20260910T064154-bada0b",
 "raw_jsonl_url": "https://licenses.ai-roundtable.de/r2/runs/run-20260910T064154-bada0b/raw.jsonl",
 "progress_jsonl_url": "https://licenses.ai-roundtable.de/r2/runs/run-20260910T064154-bada0b/progress.jsonl",
 "summary": {
  "run_id": "run-20260910T064154-bada0b",
  "question_bank_version": "2026-09-09-v5",
  "code_version": "0.2.13",
  "official": true,
  "started_at": 1789022514.8059044,
  "judges_active": [
   "gpt-5",
   "claude-opus-4-8",
   "claude-opus-4-7",
   "mistral-large-2"
  ],
  "by_model_domain": [
   {
    "model_id": "claude-fable-5-1",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.9074007936507936,
    "score_ci_lo": 0.8799603174603174,
    "score_ci_hi": 0.9336507936507936,
    "top100": 91,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.9074007936507936,
    "hallu_rate_mean": 1.6857142857142857,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.5991490053111005,
    "contradicted_rate_mean": 0.02857142857142857,
    "extra_claims_rate_mean": 1.6571428571428573,
    "extra_false_rate_mean": 0.5642857142857143,
    "extra_unsupported_rate_mean": 1.0928571428571427,
    "hallu_top100": 57,
    "inter_rater_agreement": 0.8689002267573696,
    "n_judges": 4,
    "score_mean_ex_family": 0.898344671201814,
    "self_judge_delta": 0.009056122448979576,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 16,
     "paragraph": 125,
     "study": 0,
     "number": 2,
     "other": 47
    },
    "tokens_in_total": 22926,
    "tokens_out_total": 98473,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 41,
    "duration_p90_s": 56.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-fable-5-1",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.8767625661375662,
    "score_ci_lo": 0.8290740740740741,
    "score_ci_hi": 0.9164417989417989,
    "top100": 88,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8767625661375662,
    "hallu_rate_mean": 1.5,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.7092422140965831,
    "contradicted_rate_mean": 0.16666666666666666,
    "extra_claims_rate_mean": 1.3333333333333333,
    "extra_false_rate_mean": 0.30833333333333335,
    "extra_unsupported_rate_mean": 1.025,
    "hallu_top100": 62,
    "inter_rater_agreement": 0.8400132275132275,
    "n_judges": 4,
    "score_mean_ex_family": 0.8566269841269841,
    "self_judge_delta": 0.020135582010582076,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 51,
     "other": 91
    },
    "tokens_in_total": 17883,
    "tokens_out_total": 63448,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 35,
    "duration_p90_s": 47.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-fable-5-1",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.8091674230787134,
    "score_ci_lo": 0.740820648885165,
    "score_ci_hi": 0.866442652329749,
    "top100": 81,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.8091674230787134,
    "hallu_rate_mean": 4,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.6659148273454703,
    "contradicted_rate_mean": 0.3790322580645161,
    "extra_claims_rate_mean": 3.620967741935484,
    "extra_false_rate_mean": 1.7258064516129032,
    "extra_unsupported_rate_mean": 1.8951612903225807,
    "hallu_top100": 0,
    "inter_rater_agreement": 0.8319508448540707,
    "n_judges": 4,
    "score_mean_ex_family": 0.7878956616859842,
    "self_judge_delta": 0.021271761392729127,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 11,
     "paragraph": 223,
     "study": 0,
     "number": 44,
     "other": 94
    },
    "tokens_in_total": 20250,
    "tokens_out_total": 186210,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 77.4,
    "duration_p90_s": 113.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-fable-5-1",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.7334093915343914,
    "score_ci_lo": 0.6546064814814815,
    "score_ci_hi": 0.8086805555555556,
    "top100": 73,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7334093915343914,
    "hallu_rate_mean": 3.7444444444444445,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.8136146460293432,
    "contradicted_rate_mean": 0.5694444444444444,
    "extra_claims_rate_mean": 3.175,
    "extra_false_rate_mean": 1.5527777777777778,
    "extra_unsupported_rate_mean": 1.6222222222222222,
    "hallu_top100": 6,
    "inter_rater_agreement": 0.7120899470899471,
    "n_judges": 4,
    "score_mean_ex_family": 0.679702380952381,
    "self_judge_delta": 0.05370701058201044,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 4,
     "paragraph": 231,
     "study": 0,
     "number": 12,
     "other": 72
    },
    "tokens_in_total": 22171,
    "tokens_out_total": 138067,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 64.8,
    "duration_p90_s": 106.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-7",
    "domain": "jura",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.7913775510204082,
    "score_ci_lo": 0.7303174603174603,
    "score_ci_hi": 0.8492261904761905,
    "top100": 79,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.7913775510204082,
    "hallu_rate_mean": 1.3071428571428572,
    "hallu_prevalence": 0.9142857142857143,
    "hallu_per_1k_tokens": 1.4779996123279704,
    "contradicted_rate_mean": 0.2,
    "extra_claims_rate_mean": 1.1071428571428572,
    "extra_false_rate_mean": 0.6285714285714286,
    "extra_unsupported_rate_mean": 0.4785714285714286,
    "hallu_top100": 67,
    "inter_rater_agreement": 0.8209297052154194,
    "n_judges": 4,
    "score_mean_ex_family": 0.8017970521541951,
    "self_judge_delta": -0.010419501133786868,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 2,
     "paragraph": 90,
     "study": 0,
     "number": 5,
     "other": 29
    },
    "tokens_in_total": 23031,
    "tokens_out_total": 30954,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 15.5,
    "duration_p90_s": 19.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-7",
    "domain": "medizin",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.8509160052910053,
    "score_ci_lo": 0.803670634920635,
    "score_ci_hi": 0.8937037037037037,
    "top100": 85,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8509160052910053,
    "hallu_rate_mean": 1.25,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 1.2481278082875686,
    "contradicted_rate_mean": 0.10833333333333334,
    "extra_claims_rate_mean": 1.1416666666666666,
    "extra_false_rate_mean": 0.26666666666666666,
    "extra_unsupported_rate_mean": 0.875,
    "hallu_top100": 68,
    "inter_rater_agreement": 0.8476455026455026,
    "n_judges": 4,
    "score_mean_ex_family": 0.822063492063492,
    "self_judge_delta": 0.028852513227513255,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 2,
     "number": 37,
     "other": 85
    },
    "tokens_in_total": 17973,
    "tokens_out_total": 30045,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 18.6,
    "duration_p90_s": 23.5,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-7",
    "domain": "steuerrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.7321847507331378,
    "score_ci_lo": 0.6933630661453242,
    "score_ci_hi": 0.7685035842293906,
    "top100": 73,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.7321847507331378,
    "hallu_rate_mean": 3.0483870967741935,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 2.3084815321477428,
    "contradicted_rate_mean": 0.3870967741935484,
    "extra_claims_rate_mean": 2.661290322580645,
    "extra_false_rate_mean": 1.4516129032258065,
    "extra_unsupported_rate_mean": 1.2096774193548387,
    "hallu_top100": 23,
    "inter_rater_agreement": 0.7766117395149653,
    "n_judges": 4,
    "score_mean_ex_family": 0.7271866126704836,
    "self_judge_delta": 0.004998138062654256,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 11,
     "paragraph": 143,
     "study": 0,
     "number": 40,
     "other": 68
    },
    "tokens_in_total": 20343,
    "tokens_out_total": 40936,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 22.5,
    "duration_p90_s": 29,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-7",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.7806084656084656,
    "score_ci_lo": 0.7346230158730158,
    "score_ci_hi": 0.8243088624338625,
    "top100": 78,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7806084656084656,
    "hallu_rate_mean": 2.9916666666666667,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 2.0591946770677985,
    "contradicted_rate_mean": 0.24166666666666667,
    "extra_claims_rate_mean": 2.75,
    "extra_false_rate_mean": 1.525,
    "extra_unsupported_rate_mean": 1.225,
    "hallu_top100": 25,
    "inter_rater_agreement": 0.7299338624338624,
    "n_judges": 4,
    "score_mean_ex_family": 0.7458068783068783,
    "self_judge_delta": 0.03480158730158733,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 10,
     "paragraph": 185,
     "study": 0,
     "number": 25,
     "other": 62
    },
    "tokens_in_total": 22261,
    "tokens_out_total": 43585,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 25.7,
    "duration_p90_s": 37.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-8",
    "domain": "jura",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.7978883219954648,
    "score_ci_lo": 0.7592375283446712,
    "score_ci_hi": 0.835249433106576,
    "top100": 80,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.7978883219954648,
    "hallu_rate_mean": 0.7357142857142858,
    "hallu_prevalence": 0.8,
    "hallu_per_1k_tokens": 0.7992178528197648,
    "contradicted_rate_mean": 0.04285714285714286,
    "extra_claims_rate_mean": 0.6928571428571428,
    "extra_false_rate_mean": 0.2785714285714286,
    "extra_unsupported_rate_mean": 0.4142857142857143,
    "hallu_top100": 81,
    "inter_rater_agreement": 0.7739455782312925,
    "n_judges": 4,
    "score_mean_ex_family": 0.809767573696145,
    "self_judge_delta": -0.011879251700680271,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 51,
     "study": 0,
     "number": 7,
     "other": 29
    },
    "tokens_in_total": 22856,
    "tokens_out_total": 32219,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 15,
    "duration_p90_s": 18.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-8",
    "domain": "medizin",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.8063624338624339,
    "score_ci_lo": 0.7477314814814815,
    "score_ci_hi": 0.861111111111111,
    "top100": 81,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8063624338624339,
    "hallu_rate_mean": 0.85,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 0.8728094194961665,
    "contradicted_rate_mean": 0.15,
    "extra_claims_rate_mean": 0.7,
    "extra_false_rate_mean": 0.1,
    "extra_unsupported_rate_mean": 0.6,
    "hallu_top100": 78,
    "inter_rater_agreement": 0.7923677248677249,
    "n_judges": 4,
    "score_mean_ex_family": 0.7773941798941799,
    "self_judge_delta": 0.028968253968253976,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 16,
     "other": 64
    },
    "tokens_in_total": 17823,
    "tokens_out_total": 29216,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 17.2,
    "duration_p90_s": 20.1,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-8",
    "domain": "steuerrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.7106107154494252,
    "score_ci_lo": 0.6589550458502071,
    "score_ci_hi": 0.7590126495368431,
    "top100": 71,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.7106107154494252,
    "hallu_rate_mean": 1.9516129032258065,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 1.3765329571568339,
    "contradicted_rate_mean": 0.27419354838709675,
    "extra_claims_rate_mean": 1.6774193548387097,
    "extra_false_rate_mean": 0.9112903225806451,
    "extra_unsupported_rate_mean": 0.7661290322580645,
    "hallu_top100": 51,
    "inter_rater_agreement": 0.7760496671786995,
    "n_judges": 4,
    "score_mean_ex_family": 0.6949041102266909,
    "self_judge_delta": 0.01570660522273437,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 1,
     "paragraph": 77,
     "study": 0,
     "number": 34,
     "other": 67
    },
    "tokens_in_total": 20188,
    "tokens_out_total": 43951,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 22.8,
    "duration_p90_s": 27.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-8",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.6732936507936508,
    "score_ci_lo": 0.6181944444444445,
    "score_ci_hi": 0.7353869047619047,
    "top100": 67,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6732936507936508,
    "hallu_rate_mean": 2.35,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 1.5578733371635658,
    "contradicted_rate_mean": 0.475,
    "extra_claims_rate_mean": 1.875,
    "extra_false_rate_mean": 1.0083333333333333,
    "extra_unsupported_rate_mean": 0.8666666666666667,
    "hallu_top100": 41,
    "inter_rater_agreement": 0.6863756613756614,
    "n_judges": 4,
    "score_mean_ex_family": 0.624378306878307,
    "self_judge_delta": 0.0489153439153438,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 136,
     "study": 0,
     "number": 10,
     "other": 57
    },
    "tokens_in_total": 22111,
    "tokens_out_total": 45254,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 24.9,
    "duration_p90_s": 33.5,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-5",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.9106632653061225,
    "score_ci_lo": 0.8756632653061225,
    "score_ci_hi": 0.9394302721088434,
    "top100": 91,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.9106632653061225,
    "hallu_rate_mean": 2.05,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.6017578878507808,
    "contradicted_rate_mean": 0.12142857142857143,
    "extra_claims_rate_mean": 1.9285714285714286,
    "extra_false_rate_mean": 0.55,
    "extra_unsupported_rate_mean": 1.3785714285714286,
    "hallu_top100": 48,
    "inter_rater_agreement": 0.8553741496598639,
    "n_judges": 4,
    "score_mean_ex_family": 0.8980839002267573,
    "self_judge_delta": 0.01257936507936519,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 10,
     "paragraph": 162,
     "study": 0,
     "number": 8,
     "other": 54
    },
    "tokens_in_total": 22856,
    "tokens_out_total": 119234,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 49.1,
    "duration_p90_s": 63.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-5",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.9200231481481481,
    "score_ci_lo": 0.8680092592592593,
    "score_ci_hi": 0.9584027777777777,
    "top100": 92,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.9200231481481481,
    "hallu_rate_mean": 2.05,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.6523676171079429,
    "contradicted_rate_mean": 0.19166666666666668,
    "extra_claims_rate_mean": 1.8583333333333334,
    "extra_false_rate_mean": 0.325,
    "extra_unsupported_rate_mean": 1.5333333333333334,
    "hallu_top100": 48,
    "inter_rater_agreement": 0.9001851851851852,
    "n_judges": 4,
    "score_mean_ex_family": 0.9076388888888889,
    "self_judge_delta": 0.012384259259259234,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 66,
     "other": 139
    },
    "tokens_in_total": 17823,
    "tokens_out_total": 94272,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 45.4,
    "duration_p90_s": 61.1,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-5",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.8899467020434763,
    "score_ci_lo": 0.8527684680910488,
    "score_ci_hi": 0.9222964087883443,
    "top100": 89,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.8899467020434763,
    "hallu_rate_mean": 4.25,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.5896648644777829,
    "contradicted_rate_mean": 0.3225806451612903,
    "extra_claims_rate_mean": 3.9274193548387095,
    "extra_false_rate_mean": 1.8387096774193548,
    "extra_unsupported_rate_mean": 2.088709677419355,
    "hallu_top100": 0,
    "inter_rater_agreement": 0.8625168738071963,
    "n_judges": 4,
    "score_mean_ex_family": 0.8579335986594051,
    "self_judge_delta": 0.03201310338407115,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 32,
     "paragraph": 238,
     "study": 0,
     "number": 72,
     "other": 76
    },
    "tokens_in_total": 20188,
    "tokens_out_total": 223432,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 91.5,
    "duration_p90_s": 135,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-5",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.7988756613756614,
    "score_ci_lo": 0.7377116402116403,
    "score_ci_hi": 0.8581746031746033,
    "top100": 80,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7988756613756614,
    "hallu_rate_mean": 3.7333333333333334,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.47434311948364366,
    "contradicted_rate_mean": 0.5083333333333333,
    "extra_claims_rate_mean": 3.225,
    "extra_false_rate_mean": 1.375,
    "extra_unsupported_rate_mean": 1.85,
    "hallu_top100": 6,
    "inter_rater_agreement": 0.7937698412698412,
    "n_judges": 4,
    "score_mean_ex_family": 0.7484060846560846,
    "self_judge_delta": 0.05046957671957675,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 8,
     "paragraph": 209,
     "study": 0,
     "number": 23,
     "other": 102
    },
    "tokens_in_total": 22111,
    "tokens_out_total": 236116,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 97.1,
    "duration_p90_s": 164.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "deepseek-v4-pro",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.6103769841269842,
    "score_ci_lo": 0.5350396825396825,
    "score_ci_hi": 0.6817261904761904,
    "top100": 61,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.6103769841269842,
    "hallu_rate_mean": 0.7357142857142858,
    "hallu_prevalence": 0.6285714285714286,
    "hallu_per_1k_tokens": 0.11004320531284321,
    "contradicted_rate_mean": 0.34285714285714286,
    "extra_claims_rate_mean": 0.39285714285714285,
    "extra_false_rate_mean": 0.18571428571428572,
    "extra_unsupported_rate_mean": 0.20714285714285716,
    "hallu_top100": 81,
    "inter_rater_agreement": 0.7232993197278912,
    "n_judges": 4,
    "score_mean_ex_family": 0.6103769841269842,
    "self_judge_delta": 0,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 21,
     "study": 0,
     "number": 3,
     "other": 19
    },
    "tokens_in_total": 16575,
    "tokens_out_total": 233999,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 118.1,
    "duration_p90_s": 263.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "deepseek-v4-pro",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.6796329365079364,
    "score_ci_lo": 0.5998611111111111,
    "score_ci_hi": 0.7420403439153439,
    "top100": 68,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6796329365079364,
    "hallu_rate_mean": 0.5333333333333333,
    "hallu_prevalence": 0.8,
    "hallu_per_1k_tokens": 0.243605359317905,
    "contradicted_rate_mean": 0.19166666666666668,
    "extra_claims_rate_mean": 0.3416666666666667,
    "extra_false_rate_mean": 0.125,
    "extra_unsupported_rate_mean": 0.21666666666666667,
    "hallu_top100": 86,
    "inter_rater_agreement": 0.7192195767195767,
    "n_judges": 4,
    "score_mean_ex_family": 0.6796329365079364,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 8,
     "other": 32
    },
    "tokens_in_total": 13028,
    "tokens_out_total": 65680,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 44.4,
    "duration_p90_s": 79.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "deepseek-v4-pro",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.48025851836335703,
    "score_ci_lo": 0.41330441511893123,
    "score_ci_hi": 0.5442402131918261,
    "top100": 48,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.48025851836335703,
    "hallu_rate_mean": 2.2983870967741935,
    "hallu_prevalence": 0.9354838709677419,
    "hallu_per_1k_tokens": 0.25702999585144565,
    "contradicted_rate_mean": 0.8467741935483871,
    "extra_claims_rate_mean": 1.4516129032258065,
    "extra_false_rate_mean": 0.967741935483871,
    "extra_unsupported_rate_mean": 0.4838709677419355,
    "hallu_top100": 42,
    "inter_rater_agreement": 0.648220686123912,
    "n_judges": 4,
    "score_mean_ex_family": 0.48025851836335703,
    "self_judge_delta": 0,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 43,
     "study": 0,
     "number": 52,
     "other": 44
    },
    "tokens_in_total": 14442,
    "tokens_out_total": 277205,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 154.9,
    "duration_p90_s": 254.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "deepseek-v4-pro",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.5910119047619048,
    "score_ci_lo": 0.5153075396825397,
    "score_ci_hi": 0.6657837301587303,
    "top100": 59,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.5910119047619048,
    "hallu_rate_mean": 1.9416666666666667,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 0.2517231693351483,
    "contradicted_rate_mean": 0.5,
    "extra_claims_rate_mean": 1.4416666666666667,
    "extra_false_rate_mean": 0.9,
    "extra_unsupported_rate_mean": 0.5416666666666666,
    "hallu_top100": 51,
    "inter_rater_agreement": 0.6271957671957672,
    "n_judges": 4,
    "score_mean_ex_family": 0.5910119047619048,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 69,
     "study": 0,
     "number": 15,
     "other": 59
    },
    "tokens_in_total": 15332,
    "tokens_out_total": 231405,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 147.5,
    "duration_p90_s": 281.1,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gemini-3-1-pro",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.5862018140589569,
    "score_ci_lo": 0.5202579365079365,
    "score_ci_hi": 0.6447732426303855,
    "top100": 59,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.5862018140589569,
    "hallu_rate_mean": 0.5785714285714286,
    "hallu_prevalence": 0.5428571428571428,
    "hallu_per_1k_tokens": 3.118263012011087,
    "contradicted_rate_mean": 0.24285714285714285,
    "extra_claims_rate_mean": 0.3357142857142857,
    "extra_false_rate_mean": 0.15714285714285714,
    "extra_unsupported_rate_mean": 0.17857142857142858,
    "hallu_top100": 85,
    "inter_rater_agreement": 0.6827097505668934,
    "n_judges": 4,
    "score_mean_ex_family": 0.5862018140589569,
    "self_judge_delta": 0,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 15,
     "study": 0,
     "number": 6,
     "other": 17
    },
    "tokens_in_total": 12953,
    "tokens_out_total": 6494,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 23.6,
    "duration_p90_s": 34.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gemini-3-1-pro",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.5962863756613757,
    "score_ci_lo": 0.5426322751322752,
    "score_ci_hi": 0.6498974867724867,
    "top100": 60,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.5962863756613757,
    "hallu_rate_mean": 0.4,
    "hallu_prevalence": 0.5666666666666667,
    "hallu_per_1k_tokens": 1.9014419267944858,
    "contradicted_rate_mean": 0.11666666666666667,
    "extra_claims_rate_mean": 0.2833333333333333,
    "extra_false_rate_mean": 0.09166666666666666,
    "extra_unsupported_rate_mean": 0.19166666666666668,
    "hallu_top100": 90,
    "inter_rater_agreement": 0.7326587301587302,
    "n_judges": 4,
    "score_mean_ex_family": 0.5962863756613757,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 3,
     "other": 28
    },
    "tokens_in_total": 10224,
    "tokens_out_total": 6311,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 21.3,
    "duration_p90_s": 30,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gemini-3-1-pro",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.48600870455709166,
    "score_ci_lo": 0.43226766513056836,
    "score_ci_hi": 0.5432219662058372,
    "top100": 49,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.48600870455709166,
    "hallu_rate_mean": 1.0403225806451613,
    "hallu_prevalence": 0.7419354838709677,
    "hallu_per_1k_tokens": 4.217340133385641,
    "contradicted_rate_mean": 0.4112903225806452,
    "extra_claims_rate_mean": 0.6290322580645161,
    "extra_false_rate_mean": 0.3790322580645161,
    "extra_unsupported_rate_mean": 0.25,
    "hallu_top100": 73,
    "inter_rater_agreement": 0.674204021784667,
    "n_judges": 4,
    "score_mean_ex_family": 0.48600870455709166,
    "self_judge_delta": 0,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 14,
     "study": 0,
     "number": 10,
     "other": 36
    },
    "tokens_in_total": 11090,
    "tokens_out_total": 7647,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 27.8,
    "duration_p90_s": 36.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gemini-3-1-pro",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean": 0.48924272486772485,
    "score_ci_lo": 0.4187962962962963,
    "score_ci_hi": 0.5568022486772487,
    "top100": 49,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.48924272486772485,
    "hallu_rate_mean": 1.5416666666666667,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 5.070160052620039,
    "contradicted_rate_mean": 0.6611111111111111,
    "extra_claims_rate_mean": 0.8805555555555556,
    "extra_false_rate_mean": 0.5166666666666667,
    "extra_unsupported_rate_mean": 0.3638888888888889,
    "hallu_top100": 61,
    "inter_rater_agreement": 0.5750925925925926,
    "n_judges": 4,
    "score_mean_ex_family": 0.48924272486772485,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 27,
     "study": 0,
     "number": 14,
     "other": 45
    },
    "tokens_in_total": 11544,
    "tokens_out_total": 9122,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 27.2,
    "duration_p90_s": 37.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "glm-5.3",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.7036892361111111,
    "score_ci_lo": 0.5969184027777777,
    "score_ci_hi": 0.7900607638888889,
    "top100": 70,
    "n_cells": 32,
    "closed_accuracy": null,
    "open_rubric_score": 0.7036892361111111,
    "hallu_rate_mean": 2.7421875,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.32208808512668796,
    "contradicted_rate_mean": 0.7109375,
    "extra_claims_rate_mean": 2.03125,
    "extra_false_rate_mean": 1.0859375,
    "extra_unsupported_rate_mean": 0.9453125,
    "hallu_top100": 31,
    "inter_rater_agreement": 0.7931423611111111,
    "n_judges": 4,
    "score_mean_ex_family": 0.7036892361111111,
    "self_judge_delta": 0,
    "n_cells_ex_family": 32,
    "false_claim_types": {
     "az": 4,
     "paragraph": 148,
     "study": 0,
     "number": 9,
     "other": 57
    },
    "tokens_in_total": 12069,
    "tokens_out_total": 272441,
    "n_failed": 3,
    "n_timeout": 2,
    "duration_median_s": 148.4,
    "duration_p90_s": 482.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "glm-5.3",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.8677281746031745,
    "score_ci_lo": 0.819212962962963,
    "score_ci_hi": 0.907691798941799,
    "top100": 87,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8677281746031745,
    "hallu_rate_mean": 1.5916666666666666,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.34690941849989826,
    "contradicted_rate_mean": 0.18333333333333332,
    "extra_claims_rate_mean": 1.4083333333333334,
    "extra_false_rate_mean": 0.44166666666666665,
    "extra_unsupported_rate_mean": 0.9666666666666667,
    "hallu_top100": 60,
    "inter_rater_agreement": 0.7932142857142856,
    "n_judges": 4,
    "score_mean_ex_family": 0.8677281746031745,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 51,
     "other": 101
    },
    "tokens_in_total": 10855,
    "tokens_out_total": 137644,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 83.2,
    "duration_p90_s": 157.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "glm-5.3",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.5850308641975308,
    "score_ci_lo": 0.4219907407407407,
    "score_ci_hi": 0.7095679012345679,
    "top100": 59,
    "n_cells": 9,
    "closed_accuracy": null,
    "open_rubric_score": 0.5850308641975308,
    "hallu_rate_mean": 5.083333333333333,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.5374828181722059,
    "contradicted_rate_mean": 1.2777777777777777,
    "extra_claims_rate_mean": 3.8055555555555554,
    "extra_false_rate_mean": 2.9722222222222223,
    "extra_unsupported_rate_mean": 0.8333333333333334,
    "hallu_top100": 0,
    "inter_rater_agreement": 0.7266273849607183,
    "n_judges": 4,
    "score_mean_ex_family": 0.5850308641975308,
    "self_judge_delta": 0,
    "n_cells_ex_family": 9,
    "false_claim_types": {
     "az": 13,
     "paragraph": 43,
     "study": 0,
     "number": 19,
     "other": 34
    },
    "tokens_in_total": 3054,
    "tokens_out_total": 85119,
    "n_failed": 22,
    "n_timeout": 16,
    "duration_median_s": 226.1,
    "duration_p90_s": 864.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "glm-5.3",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.6755140692640693,
    "score_ci_lo": 0.5471861471861472,
    "score_ci_hi": 0.78125,
    "top100": 68,
    "n_cells": 11,
    "closed_accuracy": null,
    "open_rubric_score": 0.6755140692640693,
    "hallu_rate_mean": 4.204545454545454,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.35317475468672443,
    "contradicted_rate_mean": 0.6363636363636364,
    "extra_claims_rate_mean": 3.5681818181818183,
    "extra_false_rate_mean": 2.3636363636363638,
    "extra_unsupported_rate_mean": 1.2045454545454546,
    "hallu_top100": 0,
    "inter_rater_agreement": 0.728030303030303,
    "n_judges": 4,
    "score_mean_ex_family": 0.6755140692640693,
    "self_judge_delta": 0,
    "n_cells_ex_family": 11,
    "false_claim_types": {
     "az": 1,
     "paragraph": 96,
     "study": 0,
     "number": 7,
     "other": 18
    },
    "tokens_in_total": 4133,
    "tokens_out_total": 130955,
    "n_failed": 19,
    "n_timeout": 6,
    "duration_median_s": 244,
    "duration_p90_s": 501,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7691213151927437,
    "score_ci_lo": 0.7040419501133787,
    "score_ci_hi": 0.832687074829932,
    "top100": 77,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.7691213151927437,
    "hallu_rate_mean": 1.2,
    "hallu_prevalence": 0.8571428571428571,
    "hallu_per_1k_tokens": 0.2814070351758794,
    "contradicted_rate_mean": 0.22142857142857142,
    "extra_claims_rate_mean": 0.9785714285714285,
    "extra_false_rate_mean": 0.42142857142857143,
    "extra_unsupported_rate_mean": 0.5571428571428572,
    "hallu_top100": 70,
    "inter_rater_agreement": 0.7687755102040816,
    "n_judges": 4,
    "score_mean_ex_family": 0.7741836734693878,
    "self_judge_delta": -0.00506235827664403,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 2,
     "paragraph": 65,
     "study": 0,
     "number": 6,
     "other": 40
    },
    "tokens_in_total": 11892,
    "tokens_out_total": 149250,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 62.9,
    "duration_p90_s": 98,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.8404431216931217,
    "score_ci_lo": 0.7750033068783069,
    "score_ci_hi": 0.8958630952380953,
    "top100": 84,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8404431216931217,
    "hallu_rate_mean": 1.3,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.407255411797876,
    "contradicted_rate_mean": 0.25833333333333336,
    "extra_claims_rate_mean": 1.0416666666666667,
    "extra_false_rate_mean": 0.19166666666666668,
    "extra_unsupported_rate_mean": 0.85,
    "hallu_top100": 67,
    "inter_rater_agreement": 0.8324206349206349,
    "n_judges": 4,
    "score_mean_ex_family": 0.8704629629629629,
    "self_judge_delta": -0.03001984126984114,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 56,
     "other": 60
    },
    "tokens_in_total": 9789,
    "tokens_out_total": 95763,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 47.5,
    "duration_p90_s": 68.4,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.6502202322766839,
    "score_ci_lo": 0.5660580458967556,
    "score_ci_hi": 0.733786016850533,
    "top100": 65,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.6502202322766839,
    "hallu_rate_mean": 2.725806451612903,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.4467259481692166,
    "contradicted_rate_mean": 0.7661290322580645,
    "extra_claims_rate_mean": 1.9596774193548387,
    "extra_false_rate_mean": 1.1290322580645162,
    "extra_unsupported_rate_mean": 0.8306451612903226,
    "hallu_top100": 31,
    "inter_rater_agreement": 0.7209642508029604,
    "n_judges": 4,
    "score_mean_ex_family": 0.677467066983196,
    "self_judge_delta": -0.027246834706512124,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 1,
     "paragraph": 83,
     "study": 0,
     "number": 63,
     "other": 50
    },
    "tokens_in_total": 10385,
    "tokens_out_total": 189154,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 81.8,
    "duration_p90_s": 196.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7965806878306878,
    "score_ci_lo": 0.721468253968254,
    "score_ci_hi": 0.8598048941798943,
    "top100": 80,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7965806878306878,
    "hallu_rate_mean": 1.825,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 0.33019123952886686,
    "contradicted_rate_mean": 0.3416666666666667,
    "extra_claims_rate_mean": 1.4833333333333334,
    "extra_false_rate_mean": 0.5916666666666667,
    "extra_unsupported_rate_mean": 0.8916666666666667,
    "hallu_top100": 54,
    "inter_rater_agreement": 0.7485449735449735,
    "n_judges": 4,
    "score_mean_ex_family": 0.8354938271604938,
    "self_judge_delta": -0.038913139329806,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 82,
     "study": 0,
     "number": 15,
     "other": 60
    },
    "tokens_in_total": 11128,
    "tokens_out_total": 165813,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 97.3,
    "duration_p90_s": 155.7,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5.6-sol",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7746995464852607,
    "score_ci_lo": 0.71906179138322,
    "score_ci_hi": 0.8182539682539682,
    "top100": 77,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.7746995464852607,
    "hallu_rate_mean": 0.9214285714285714,
    "hallu_prevalence": 0.8285714285714286,
    "hallu_per_1k_tokens": 0.5453069782384471,
    "contradicted_rate_mean": 0.12857142857142856,
    "extra_claims_rate_mean": 0.7928571428571428,
    "extra_false_rate_mean": 0.3,
    "extra_unsupported_rate_mean": 0.4928571428571429,
    "hallu_top100": 76,
    "inter_rater_agreement": 0.7580725623582766,
    "n_judges": 4,
    "score_mean_ex_family": 0.7944217687074829,
    "self_judge_delta": -0.019722222222222197,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 68,
     "study": 0,
     "number": 4,
     "other": 17
    },
    "tokens_in_total": 11892,
    "tokens_out_total": 59141,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 45.3,
    "duration_p90_s": 85.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5.6-sol",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.8326951058201059,
    "score_ci_lo": 0.7862632275132275,
    "score_ci_hi": 0.8765013227513226,
    "top100": 83,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8326951058201059,
    "hallu_rate_mean": 0.75,
    "hallu_prevalence": 0.9,
    "hallu_per_1k_tokens": 0.7755678880424666,
    "contradicted_rate_mean": 0.175,
    "extra_claims_rate_mean": 0.575,
    "extra_false_rate_mean": 0.125,
    "extra_unsupported_rate_mean": 0.45,
    "hallu_top100": 81,
    "inter_rater_agreement": 0.7788095238095238,
    "n_judges": 4,
    "score_mean_ex_family": 0.8774911816578483,
    "self_judge_delta": -0.04479607583774248,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 16,
     "other": 46
    },
    "tokens_in_total": 9789,
    "tokens_out_total": 29011,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 26.2,
    "duration_p90_s": 53.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5.6-sol",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7316765000232742,
    "score_ci_lo": 0.6751504096262161,
    "score_ci_hi": 0.780130801098543,
    "top100": 73,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.7316765000232742,
    "hallu_rate_mean": 1.7741935483870968,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.524173949507753,
    "contradicted_rate_mean": 0.3064516129032258,
    "extra_claims_rate_mean": 1.467741935483871,
    "extra_false_rate_mean": 0.7258064516129032,
    "extra_unsupported_rate_mean": 0.7419354838709677,
    "hallu_top100": 55,
    "inter_rater_agreement": 0.780657962109575,
    "n_judges": 4,
    "score_mean_ex_family": 0.7575800245155084,
    "self_judge_delta": -0.02590352449223421,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 64,
     "study": 0,
     "number": 40,
     "other": 49
    },
    "tokens_in_total": 10385,
    "tokens_out_total": 104927,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 90.4,
    "duration_p90_s": 133.5,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5.6-sol",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7791997354497354,
    "score_ci_lo": 0.7217195767195768,
    "score_ci_hi": 0.8358862433862434,
    "top100": 78,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7791997354497354,
    "hallu_rate_mean": 1.4666666666666666,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.557145389621901,
    "contradicted_rate_mean": 0.3416666666666667,
    "extra_claims_rate_mean": 1.125,
    "extra_false_rate_mean": 0.5833333333333334,
    "extra_unsupported_rate_mean": 0.5416666666666666,
    "hallu_top100": 63,
    "inter_rater_agreement": 0.7302380952380952,
    "n_judges": 4,
    "score_mean_ex_family": 0.8321957671957672,
    "self_judge_delta": -0.052996031746031735,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 55,
     "study": 0,
     "number": 10,
     "other": 57
    },
    "tokens_in_total": 11128,
    "tokens_out_total": 78974,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 76.9,
    "duration_p90_s": 136.7,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-6-astra",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7343820861678004,
    "score_ci_lo": 0.6677380952380952,
    "score_ci_hi": 0.7941468253968255,
    "top100": 73,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.7343820861678004,
    "hallu_rate_mean": 1.2857142857142858,
    "hallu_prevalence": 0.9142857142857143,
    "hallu_per_1k_tokens": 1.2257572455872738,
    "contradicted_rate_mean": 0.2571428571428571,
    "extra_claims_rate_mean": 1.0285714285714285,
    "extra_false_rate_mean": 0.5071428571428571,
    "extra_unsupported_rate_mean": 0.5214285714285715,
    "hallu_top100": 67,
    "inter_rater_agreement": 0.6884353741496598,
    "n_judges": 4,
    "score_mean_ex_family": 0.7718820861678004,
    "self_judge_delta": -0.03749999999999998,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 82,
     "study": 0,
     "number": 13,
     "other": 30
    },
    "tokens_in_total": 11892,
    "tokens_out_total": 36712,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 33.8,
    "duration_p90_s": 56.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-6-astra",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.81676917989418,
    "score_ci_lo": 0.7808597883597883,
    "score_ci_hi": 0.8532539682539683,
    "top100": 82,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.81676917989418,
    "hallu_rate_mean": 0.9666666666666667,
    "hallu_prevalence": 0.9,
    "hallu_per_1k_tokens": 1.1398474962660168,
    "contradicted_rate_mean": 0.26666666666666666,
    "extra_claims_rate_mean": 0.7,
    "extra_false_rate_mean": 0.175,
    "extra_unsupported_rate_mean": 0.525,
    "hallu_top100": 75,
    "inter_rater_agreement": 0.7779761904761905,
    "n_judges": 4,
    "score_mean_ex_family": 0.8605776014109348,
    "self_judge_delta": -0.04380842151675479,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 21,
     "other": 57
    },
    "tokens_in_total": 9789,
    "tokens_out_total": 25442,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 26.8,
    "duration_p90_s": 37.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-6-astra",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.736117220593027,
    "score_ci_lo": 0.6750692407950473,
    "score_ci_hi": 0.7930840664711633,
    "top100": 74,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.736117220593027,
    "hallu_rate_mean": 1.8790322580645162,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.9758426589827782,
    "contradicted_rate_mean": 0.31451612903225806,
    "extra_claims_rate_mean": 1.564516129032258,
    "extra_false_rate_mean": 0.6612903225806451,
    "extra_unsupported_rate_mean": 0.9032258064516129,
    "hallu_top100": 53,
    "inter_rater_agreement": 0.7220150816925011,
    "n_judges": 4,
    "score_mean_ex_family": 0.7723591521978619,
    "self_judge_delta": -0.03624193160483491,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 43,
     "study": 0,
     "number": 50,
     "other": 76
    },
    "tokens_in_total": 10385,
    "tokens_out_total": 59692,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 57.1,
    "duration_p90_s": 85.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-6-astra",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean": 0.7544775132275132,
    "score_ci_lo": 0.6872718253968254,
    "score_ci_hi": 0.824785052910053,
    "top100": 75,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7544775132275132,
    "hallu_rate_mean": 1.575,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 0.9308143887159687,
    "contradicted_rate_mean": 0.3333333333333333,
    "extra_claims_rate_mean": 1.2416666666666667,
    "extra_false_rate_mean": 0.6333333333333333,
    "extra_unsupported_rate_mean": 0.6083333333333333,
    "hallu_top100": 60,
    "inter_rater_agreement": 0.7002777777777778,
    "n_judges": 4,
    "score_mean_ex_family": 0.8068121693121693,
    "self_judge_delta": -0.05233465608465604,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 63,
     "study": 0,
     "number": 17,
     "other": 59
    },
    "tokens_in_total": 11128,
    "tokens_out_total": 50762,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 56.8,
    "duration_p90_s": 94.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.5",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.652562358276644,
    "score_ci_lo": 0.5891496598639455,
    "score_ci_hi": 0.7033985260770975,
    "top100": 65,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.652562358276644,
    "hallu_rate_mean": 0.5642857142857143,
    "hallu_prevalence": 0.5142857142857142,
    "hallu_per_1k_tokens": 2.8364210828665803,
    "contradicted_rate_mean": 0.20714285714285716,
    "extra_claims_rate_mean": 0.35714285714285715,
    "extra_false_rate_mean": 0.2,
    "extra_unsupported_rate_mean": 0.15714285714285714,
    "hallu_top100": 85,
    "inter_rater_agreement": 0.7015192743764173,
    "n_judges": 4,
    "score_mean_ex_family": 0.652562358276644,
    "self_judge_delta": 0,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 22,
     "study": 0,
     "number": 3,
     "other": 14
    },
    "tokens_in_total": 28950,
    "tokens_out_total": 6963,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 30.5,
    "duration_p90_s": 45,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.5",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.7016798941798942,
    "score_ci_lo": 0.6465674603174603,
    "score_ci_hi": 0.7590906084656085,
    "top100": 70,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.7016798941798942,
    "hallu_rate_mean": 0.45,
    "hallu_prevalence": 0.7666666666666667,
    "hallu_per_1k_tokens": 1.5919811320754718,
    "contradicted_rate_mean": 0.13333333333333333,
    "extra_claims_rate_mean": 0.31666666666666665,
    "extra_false_rate_mean": 0.058333333333333334,
    "extra_unsupported_rate_mean": 0.25833333333333336,
    "hallu_top100": 88,
    "inter_rater_agreement": 0.7005423280423281,
    "n_judges": 4,
    "score_mean_ex_family": 0.7016798941798942,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 7,
     "other": 28
    },
    "tokens_in_total": 24284,
    "tokens_out_total": 8480,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 20.4,
    "duration_p90_s": 27.4,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.5",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.5796818996415771,
    "score_ci_lo": 0.5270833333333333,
    "score_ci_hi": 0.6291890681003585,
    "top100": 58,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.5796818996415771,
    "hallu_rate_mean": 1.25,
    "hallu_prevalence": 0.967741935483871,
    "hallu_per_1k_tokens": 3.613727501632006,
    "contradicted_rate_mean": 0.3548387096774194,
    "extra_claims_rate_mean": 0.8951612903225806,
    "extra_false_rate_mean": 0.4596774193548387,
    "extra_unsupported_rate_mean": 0.43548387096774194,
    "hallu_top100": 68,
    "inter_rater_agreement": 0.7045547642321837,
    "n_judges": 4,
    "score_mean_ex_family": 0.5796818996415771,
    "self_judge_delta": 0,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 29,
     "study": 0,
     "number": 9,
     "other": 56
    },
    "tokens_in_total": 25335,
    "tokens_out_total": 10723,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 40.9,
    "duration_p90_s": 52.9,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.5",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.6658498677248678,
    "score_ci_lo": 0.6113293650793651,
    "score_ci_hi": 0.7097420634920635,
    "top100": 67,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6658498677248678,
    "hallu_rate_mean": 1,
    "hallu_prevalence": 0.9,
    "hallu_per_1k_tokens": 2.279981760145919,
    "contradicted_rate_mean": 0.225,
    "extra_claims_rate_mean": 0.775,
    "extra_false_rate_mean": 0.30833333333333335,
    "extra_unsupported_rate_mean": 0.4666666666666667,
    "hallu_top100": 75,
    "inter_rater_agreement": 0.6003042328042327,
    "n_judges": 4,
    "score_mean_ex_family": 0.6658498677248678,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 31,
     "study": 0,
     "number": 7,
     "other": 44
    },
    "tokens_in_total": 25514,
    "tokens_out_total": 13158,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 29.6,
    "duration_p90_s": 45.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.6",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.6437018140589569,
    "score_ci_lo": 0.5582341269841269,
    "score_ci_hi": 0.7121626984126984,
    "top100": 64,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.6437018140589569,
    "hallu_rate_mean": 0.65,
    "hallu_prevalence": 0.5142857142857142,
    "hallu_per_1k_tokens": 3.360413589364845,
    "contradicted_rate_mean": 0.2857142857142857,
    "extra_claims_rate_mean": 0.36428571428571427,
    "extra_false_rate_mean": 0.22857142857142856,
    "extra_unsupported_rate_mean": 0.1357142857142857,
    "hallu_top100": 83,
    "inter_rater_agreement": 0.7156462585034014,
    "n_judges": 4,
    "score_mean_ex_family": 0.6437018140589569,
    "self_judge_delta": 0,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 18,
     "study": 0,
     "number": 3,
     "other": 20
    },
    "tokens_in_total": 33920,
    "tokens_out_total": 6770,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 60.9,
    "duration_p90_s": 106.5,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.6",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.6444246031746032,
    "score_ci_lo": 0.5780357142857143,
    "score_ci_hi": 0.7027579365079365,
    "top100": 64,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6444246031746032,
    "hallu_rate_mean": 0.5583333333333333,
    "hallu_prevalence": 0.7333333333333333,
    "hallu_per_1k_tokens": 2.4156331122007497,
    "contradicted_rate_mean": 0.15833333333333333,
    "extra_claims_rate_mean": 0.4,
    "extra_false_rate_mean": 0.075,
    "extra_unsupported_rate_mean": 0.325,
    "hallu_top100": 86,
    "inter_rater_agreement": 0.7429365079365079,
    "n_judges": 4,
    "score_mean_ex_family": 0.6444246031746032,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 0,
     "number": 6,
     "other": 38
    },
    "tokens_in_total": 28544,
    "tokens_out_total": 6934,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 38.7,
    "duration_p90_s": 59,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.6",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.5252894730717311,
    "score_ci_lo": 0.46132697947214074,
    "score_ci_hi": 0.5838002723083369,
    "top100": 53,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.5252894730717311,
    "hallu_rate_mean": 1.9112903225806452,
    "hallu_prevalence": 0.967741935483871,
    "hallu_per_1k_tokens": 4.723750298971538,
    "contradicted_rate_mean": 0.6451612903225806,
    "extra_claims_rate_mean": 1.2661290322580645,
    "extra_false_rate_mean": 0.7419354838709677,
    "extra_unsupported_rate_mean": 0.5241935483870968,
    "hallu_top100": 52,
    "inter_rater_agreement": 0.6860587441232603,
    "n_judges": 4,
    "score_mean_ex_family": 0.5252894730717311,
    "self_judge_delta": 0,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 60,
     "study": 0,
     "number": 16,
     "other": 55
    },
    "tokens_in_total": 29737,
    "tokens_out_total": 12543,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 93.4,
    "duration_p90_s": 173.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.6",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean": 0.6898148148148149,
    "score_ci_lo": 0.6242890211640212,
    "score_ci_hi": 0.7511607142857143,
    "top100": 69,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6898148148148149,
    "hallu_rate_mean": 1.4166666666666667,
    "hallu_prevalence": 0.9666666666666667,
    "hallu_per_1k_tokens": 2.7708958143173814,
    "contradicted_rate_mean": 0.375,
    "extra_claims_rate_mean": 1.0416666666666667,
    "extra_false_rate_mean": 0.5333333333333333,
    "extra_unsupported_rate_mean": 0.5083333333333333,
    "hallu_top100": 64,
    "inter_rater_agreement": 0.659589947089947,
    "n_judges": 4,
    "score_mean_ex_family": 0.6898148148148149,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 58,
     "study": 0,
     "number": 4,
     "other": 50
    },
    "tokens_in_total": 29774,
    "tokens_out_total": 15338,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 76.8,
    "duration_p90_s": 128.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "kimi-k3",
    "domain": "jura",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.8114408263305322,
    "score_ci_lo": 0.7615254435107376,
    "score_ci_hi": 0.8600898692810458,
    "top100": 81,
    "n_cells": 34,
    "closed_accuracy": null,
    "open_rubric_score": 0.8114408263305322,
    "hallu_rate_mean": 1.9705882352941178,
    "hallu_prevalence": 0.8823529411764706,
    "hallu_per_1k_tokens": 0.3951869765247139,
    "contradicted_rate_mean": 0.2377450980392157,
    "extra_claims_rate_mean": 1.7328431372549018,
    "extra_false_rate_mean": 0.9191176470588235,
    "extra_unsupported_rate_mean": 0.8137254901960784,
    "hallu_top100": 50,
    "inter_rater_agreement": 0.7906979458450047,
    "n_judges": 4,
    "score_mean_ex_family": 0.8114408263305322,
    "self_judge_delta": 0,
    "n_cells_ex_family": 34,
    "false_claim_types": {
     "az": 3,
     "paragraph": 150,
     "study": 0,
     "number": 6,
     "other": 38
    },
    "tokens_in_total": 17543,
    "tokens_out_total": 169540,
    "n_failed": 1,
    "n_timeout": 1,
    "duration_median_s": 143.4,
    "duration_p90_s": 243.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "kimi-k3",
    "domain": "medizin",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.8658796296296296,
    "score_ci_lo": 0.8127314814814814,
    "score_ci_hi": 0.9106018518518518,
    "top100": 87,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.8658796296296296,
    "hallu_rate_mean": 1.5583333333333333,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.4993644452514981,
    "contradicted_rate_mean": 0.225,
    "extra_claims_rate_mean": 1.3333333333333333,
    "extra_false_rate_mean": 0.225,
    "extra_unsupported_rate_mean": 1.1083333333333334,
    "hallu_top100": 61,
    "inter_rater_agreement": 0.836574074074074,
    "n_judges": 4,
    "score_mean_ex_family": 0.8658796296296296,
    "self_judge_delta": 0,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 50,
     "other": 91
    },
    "tokens_in_total": 14166,
    "tokens_out_total": 93619,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 88.1,
    "duration_p90_s": 142.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "kimi-k3",
    "domain": "steuerrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.5776292688792689,
    "score_ci_lo": 0.4326785714285714,
    "score_ci_hi": 0.7121464646464646,
    "top100": 58,
    "n_cells": 15,
    "closed_accuracy": null,
    "open_rubric_score": 0.5776292688792689,
    "hallu_rate_mean": 4.216666666666667,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.4896041366711563,
    "contradicted_rate_mean": 1.1666666666666667,
    "extra_claims_rate_mean": 3.05,
    "extra_false_rate_mean": 1.9166666666666667,
    "extra_unsupported_rate_mean": 1.1333333333333333,
    "hallu_top100": 0,
    "inter_rater_agreement": 0.7193650793650794,
    "n_judges": 4,
    "score_mean_ex_family": 0.5776292688792689,
    "self_judge_delta": 0,
    "n_cells_ex_family": 15,
    "false_claim_types": {
     "az": 6,
     "paragraph": 71,
     "study": 0,
     "number": 35,
     "other": 36
    },
    "tokens_in_total": 7520,
    "tokens_out_total": 129186,
    "n_failed": 16,
    "n_timeout": 16,
    "duration_median_s": 246,
    "duration_p90_s": 572,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "kimi-k3",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean": 0.738558201058201,
    "score_ci_lo": 0.652980324074074,
    "score_ci_hi": 0.8154513888888889,
    "top100": 74,
    "n_cells": 24,
    "closed_accuracy": null,
    "open_rubric_score": 0.738558201058201,
    "hallu_rate_mean": 3.59375,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 0.5029301145806001,
    "contradicted_rate_mean": 0.5208333333333334,
    "extra_claims_rate_mean": 3.0729166666666665,
    "extra_false_rate_mean": 1.875,
    "extra_unsupported_rate_mean": 1.1979166666666667,
    "hallu_top100": 10,
    "inter_rater_agreement": 0.7278935185185186,
    "n_judges": 4,
    "score_mean_ex_family": 0.738558201058201,
    "self_judge_delta": 0,
    "n_cells_ex_family": 24,
    "false_claim_types": {
     "az": 0,
     "paragraph": 176,
     "study": 0,
     "number": 17,
     "other": 66
    },
    "tokens_in_total": 12966,
    "tokens_out_total": 171495,
    "n_failed": 6,
    "n_timeout": 6,
    "duration_median_s": 215.8,
    "duration_p90_s": 274,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "mistral-large-2",
    "domain": "jura",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.4273044217687075,
    "score_ci_lo": 0.3487131519274376,
    "score_ci_hi": 0.5079790249433107,
    "top100": 43,
    "n_cells": 35,
    "closed_accuracy": null,
    "open_rubric_score": 0.4273044217687075,
    "hallu_rate_mean": 2.05,
    "hallu_prevalence": 0.8857142857142857,
    "hallu_per_1k_tokens": 7.147838214783821,
    "contradicted_rate_mean": 0.9714285714285714,
    "extra_claims_rate_mean": 1.0785714285714285,
    "extra_false_rate_mean": 0.8357142857142857,
    "extra_unsupported_rate_mean": 0.24285714285714285,
    "hallu_top100": 48,
    "inter_rater_agreement": 0.6412018140589569,
    "n_judges": 4,
    "score_mean_ex_family": 0.3830801209372638,
    "self_judge_delta": 0.044224300831443686,
    "n_cells_ex_family": 35,
    "false_claim_types": {
     "az": 0,
     "paragraph": 58,
     "study": 0,
     "number": 3,
     "other": 47
    },
    "tokens_in_total": 13389,
    "tokens_out_total": 10038,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 4.9,
    "duration_p90_s": 64.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "mistral-large-2",
    "domain": "medizin",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.6523908730158731,
    "score_ci_lo": 0.6025859788359789,
    "score_ci_hi": 0.7019378306878307,
    "top100": 65,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.6523908730158731,
    "hallu_rate_mean": 1.4583333333333333,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 2.8549986948577395,
    "contradicted_rate_mean": 0.25833333333333336,
    "extra_claims_rate_mean": 1.2,
    "extra_false_rate_mean": 0.3,
    "extra_unsupported_rate_mean": 0.9,
    "hallu_top100": 63,
    "inter_rater_agreement": 0.693968253968254,
    "n_judges": 4,
    "score_mean_ex_family": 0.6189550264550264,
    "self_judge_delta": 0.03343584656084664,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 0,
     "study": 1,
     "number": 27,
     "other": 96
    },
    "tokens_in_total": 10461,
    "tokens_out_total": 15324,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 8.2,
    "duration_p90_s": 12.2,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "mistral-large-2",
    "domain": "steuerrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.27157956291020807,
    "score_ci_lo": 0.20617814085556022,
    "score_ci_hi": 0.3329338895871154,
    "top100": 27,
    "n_cells": 31,
    "closed_accuracy": null,
    "open_rubric_score": 0.27157956291020807,
    "hallu_rate_mean": 3.2661290322580645,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 7.222341108495613,
    "contradicted_rate_mean": 1.403225806451613,
    "extra_claims_rate_mean": 1.8629032258064515,
    "extra_false_rate_mean": 1.3951612903225807,
    "extra_unsupported_rate_mean": 0.46774193548387094,
    "hallu_top100": 18,
    "inter_rater_agreement": 0.5291416468835825,
    "n_judges": 4,
    "score_mean_ex_family": 0.22399804496578687,
    "self_judge_delta": 0.0475815179444212,
    "n_cells_ex_family": 31,
    "false_claim_types": {
     "az": 0,
     "paragraph": 60,
     "study": 0,
     "number": 38,
     "other": 86
    },
    "tokens_in_total": 11379,
    "tokens_out_total": 14019,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 7.6,
    "duration_p90_s": 10.1,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "mistral-large-2",
    "domain": "wirtschaftsrecht",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean": 0.5887962962962963,
    "score_ci_lo": 0.5191269841269841,
    "score_ci_hi": 0.6550099206349207,
    "top100": 59,
    "n_cells": 30,
    "closed_accuracy": null,
    "open_rubric_score": 0.5887962962962963,
    "hallu_rate_mean": 2.441666666666667,
    "hallu_prevalence": 1,
    "hallu_per_1k_tokens": 3.9263507718696395,
    "contradicted_rate_mean": 0.35833333333333334,
    "extra_claims_rate_mean": 2.0833333333333335,
    "extra_false_rate_mean": 1.3666666666666667,
    "extra_unsupported_rate_mean": 0.7166666666666667,
    "hallu_top100": 38,
    "inter_rater_agreement": 0.5571031746031746,
    "n_judges": 4,
    "score_mean_ex_family": 0.5389638447971782,
    "self_judge_delta": 0.04983245149911808,
    "n_cells_ex_family": 30,
    "false_claim_types": {
     "az": 0,
     "paragraph": 102,
     "study": 0,
     "number": 25,
     "other": 77
    },
    "tokens_in_total": 11869,
    "tokens_out_total": 18656,
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 10.7,
    "duration_p90_s": 18.7,
    "reliability_source": "derived-from-raw"
   }
  ],
  "by_model_overall": [
   {
    "model_id": "claude-fable-5-1",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.8345109255823542,
    "score_ci_lo": 0.8036186696900982,
    "score_ci_hi": 0.8626193911908198,
    "top100": 83,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 2.701058201058201,
    "hallu_top100_across_domains": 32,
    "hallu_per_1k_tokens": 0.6999891676504908,
    "contradicted_rate_mean": 0.27645502645502645,
    "extra_claims_rate_mean": 2.4246031746031744,
    "extra_false_rate_mean": 1.0244708994708995,
    "extra_unsupported_rate_mean": 1.4001322751322751,
    "score_mean_ex_family": 0.8091802377516664,
    "self_judge_delta": 0.025330687830687815,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 31,
     "paragraph": 579,
     "study": 1,
     "number": 109,
     "other": 304
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 49.2,
    "duration_p90_s": 91.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-7",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.7884259975033785,
    "score_ci_lo": 0.7623575608694657,
    "score_ci_hi": 0.8135047327011613,
    "top100": 79,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 2.123015873015873,
    "hallu_top100_across_domains": 46,
    "hallu_per_1k_tokens": 1.838235294117647,
    "contradicted_rate_mean": 0.23412698412698413,
    "extra_claims_rate_mean": 1.8888888888888888,
    "extra_false_rate_mean": 0.9583333333333334,
    "extra_unsupported_rate_mean": 0.9305555555555556,
    "score_mean_ex_family": 0.7749348645181979,
    "self_judge_delta": 0.013491132985180565,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 23,
     "paragraph": 418,
     "study": 2,
     "number": 107,
     "other": 244
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 19.3,
    "duration_p90_s": 29.4,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-4-8",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.7487675078448888,
    "score_ci_lo": 0.7210482804232805,
    "score_ci_hi": 0.7747367381295953,
    "top100": 75,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.4464285714285714,
    "hallu_top100_across_domains": 63,
    "hallu_per_1k_tokens": 1.2098380244291025,
    "contradicted_rate_mean": 0.22817460317460317,
    "extra_claims_rate_mean": 1.2182539682539681,
    "extra_false_rate_mean": 0.5654761904761905,
    "extra_unsupported_rate_mean": 0.6527777777777778,
    "score_mean_ex_family": 0.729659262694977,
    "self_judge_delta": 0.019108245149911718,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 1,
     "paragraph": 264,
     "study": 1,
     "number": 67,
     "other": 217
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 18.9,
    "duration_p90_s": 27.8,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "claude-opus-5",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.8811787804347327,
    "score_ci_lo": 0.8575373204837491,
    "score_ci_hi": 0.9029813440527726,
    "top100": 88,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 2.992063492063492,
    "hallu_top100_across_domains": 25,
    "hallu_per_1k_tokens": 0.5601333622562231,
    "contradicted_rate_mean": 0.27976190476190477,
    "extra_claims_rate_mean": 2.7123015873015874,
    "extra_false_rate_mean": 1.0099206349206349,
    "extra_unsupported_rate_mean": 1.7023809523809523,
    "score_mean_ex_family": 0.8548430735930735,
    "self_judge_delta": 0.0263357068416592,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 50,
     "paragraph": 609,
     "study": 1,
     "number": 169,
     "other": 371
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 62.9,
    "duration_p90_s": 128.5,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "deepseek-v4-pro",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.5902424900936806,
    "score_ci_lo": 0.5524864575459814,
    "score_ci_hi": 0.6268118400558876,
    "top100": 59,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.3591269841269842,
    "hallu_top100_across_domains": 66,
    "hallu_per_1k_tokens": 0.21186729004106206,
    "contradicted_rate_mean": 0.46825396825396826,
    "extra_claims_rate_mean": 0.8908730158730159,
    "extra_false_rate_mean": 0.5337301587301587,
    "extra_unsupported_rate_mean": 0.35714285714285715,
    "score_mean_ex_family": 0.5902424900936806,
    "self_judge_delta": 0,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 133,
     "study": 0,
     "number": 78,
     "other": 154
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 118.1,
    "duration_p90_s": 256.4,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gemini-3-1-pro",
    "reasoning_mode": "on",
    "reasoning_level": "adaptive",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.5408667170571932,
    "score_ci_lo": 0.5121488410178886,
    "score_ci_hi": 0.5729292642983119,
    "top100": 54,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 0.878968253968254,
    "hallu_top100_across_domains": 78,
    "hallu_per_1k_tokens": 3.7448434435652938,
    "contradicted_rate_mean": 0.35383597883597884,
    "extra_claims_rate_mean": 0.5251322751322751,
    "extra_false_rate_mean": 0.28174603174603174,
    "extra_unsupported_rate_mean": 0.2433862433862434,
    "score_mean_ex_family": 0.5408667170571932,
    "self_judge_delta": 0,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 56,
     "study": 0,
     "number": 33,
     "other": 126
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 24.1,
    "duration_p90_s": 35.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "glm-5.3",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.746900406504065,
    "score_ci_lo": 0.6975754936120789,
    "score_ci_hi": 0.7941298877274487,
    "top100": 75,
    "n_cells": 82,
    "hallu_rate_mean_across_domains": 2.774390243902439,
    "hallu_top100_across_domains": 30,
    "hallu_per_1k_tokens": 0.3633262478060684,
    "contradicted_rate_mean": 0.5701219512195121,
    "extra_claims_rate_mean": 2.2042682926829267,
    "extra_false_rate_mean": 1.228658536585366,
    "extra_unsupported_rate_mean": 0.975609756097561,
    "score_mean_ex_family": 0.746900406504065,
    "self_judge_delta": 0,
    "n_cells_ex_family": 82,
    "false_claim_types": {
     "az": 18,
     "paragraph": 287,
     "study": 1,
     "number": 86,
     "other": 210
    },
    "n_failed": 44,
    "n_timeout": 24,
    "duration_median_s": 125.8,
    "duration_p90_s": 408.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.7633872025241073,
    "score_ci_lo": 0.7245212899974806,
    "score_ci_hi": 0.8004344751368561,
    "top100": 76,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.748015873015873,
    "hallu_top100_across_domains": 56,
    "hallu_per_1k_tokens": 0.36709556985232844,
    "contradicted_rate_mean": 0.39285714285714285,
    "extra_claims_rate_mean": 1.3551587301587302,
    "extra_false_rate_mean": 0.5813492063492064,
    "extra_unsupported_rate_mean": 0.7738095238095238,
    "score_mean_ex_family": 0.7879096139016774,
    "self_judge_delta": -0.024522411377570053,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 3,
     "paragraph": 230,
     "study": 0,
     "number": 140,
     "other": 210
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 66.8,
    "duration_p90_s": 137.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-5.6-sol",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.77899445131588,
    "score_ci_lo": 0.7514158735884926,
    "score_ci_hi": 0.8045219771410247,
    "top100": 78,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.2202380952380953,
    "hallu_top100_across_domains": 69,
    "hallu_per_1k_tokens": 0.5651472323407571,
    "contradicted_rate_mean": 0.23412698412698413,
    "extra_claims_rate_mean": 0.9861111111111112,
    "extra_false_rate_mean": 0.4305555555555556,
    "extra_unsupported_rate_mean": 0.5555555555555556,
    "score_mean_ex_family": 0.8141297708758026,
    "self_judge_delta": -0.03513531955992266,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 187,
     "study": 1,
     "number": 70,
     "other": 169
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 55.8,
    "duration_p90_s": 114.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "gpt-6-astra",
    "reasoning_mode": "on",
    "reasoning_level": "medium",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.7592095845072035,
    "score_ci_lo": 0.7308137354268306,
    "score_ci_hi": 0.7875133134061706,
    "top100": 76,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.4246031746031746,
    "hallu_top100_across_domains": 64,
    "hallu_per_1k_tokens": 1.0399286243974786,
    "contradicted_rate_mean": 0.2916666666666667,
    "extra_claims_rate_mean": 1.132936507936508,
    "extra_false_rate_mean": 0.49603174603174605,
    "extra_unsupported_rate_mean": 0.6369047619047619,
    "score_mean_ex_family": 0.8014341258388877,
    "self_judge_delta": -0.042224541331684184,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 188,
     "study": 0,
     "number": 101,
     "other": 222
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 39.7,
    "duration_p90_s": 76.3,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.5",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.6494897959183674,
    "score_ci_lo": 0.6216939090450995,
    "score_ci_hi": 0.6787863756613757,
    "top100": 65,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 0.8095238095238095,
    "hallu_top100_across_domains": 79,
    "hallu_per_1k_tokens": 2.593835825450107,
    "contradicted_rate_mean": 0.23015873015873015,
    "extra_claims_rate_mean": 0.5793650793650794,
    "extra_false_rate_mean": 0.25595238095238093,
    "extra_unsupported_rate_mean": 0.32341269841269843,
    "score_mean_ex_family": 0.6494897959183674,
    "self_judge_delta": 0,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 82,
     "study": 0,
     "number": 26,
     "other": 142
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 30,
    "duration_p90_s": 48.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "grok-4.6",
    "reasoning_mode": "on",
    "reasoning_level": "high",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.6257199975949976,
    "score_ci_lo": 0.5894202798964704,
    "score_ci_hi": 0.6615236119700405,
    "top100": 63,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 1.121031746031746,
    "hallu_top100_across_domains": 71,
    "hallu_per_1k_tokens": 3.3966574485992544,
    "contradicted_rate_mean": 0.36507936507936506,
    "extra_claims_rate_mean": 0.7559523809523809,
    "extra_false_rate_mean": 0.39087301587301587,
    "extra_unsupported_rate_mean": 0.36507936507936506,
    "score_mean_ex_family": 0.6257199975949976,
    "self_judge_delta": 0,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 136,
     "study": 0,
     "number": 29,
     "other": 163
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 64.9,
    "duration_p90_s": 123.6,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "kimi-k3",
    "reasoning_mode": "on",
    "reasoning_level": "standard",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.7762642023564353,
    "score_ci_lo": 0.7354826874851147,
    "score_ci_hi": 0.8141816920943133,
    "top100": 78,
    "n_cells": 103,
    "hallu_rate_mean_across_domains": 2.5558252427184467,
    "hallu_top100_across_domains": 36,
    "hallu_per_1k_tokens": 0.4668877695800227,
    "contradicted_rate_mean": 0.4352750809061489,
    "extra_claims_rate_mean": 2.1205501618122975,
    "extra_false_rate_mean": 1.0849514563106797,
    "extra_unsupported_rate_mean": 1.0355987055016183,
    "score_mean_ex_family": 0.7762642023564353,
    "self_judge_delta": 0,
    "n_cells_ex_family": 103,
    "false_claim_types": {
     "az": 9,
     "paragraph": 397,
     "study": 1,
     "number": 108,
     "other": 231
    },
    "n_failed": 23,
    "n_timeout": 23,
    "duration_median_s": 145.5,
    "duration_p90_s": 272.1,
    "reliability_source": "derived-from-raw"
   },
   {
    "model_id": "mistral-large-2",
    "reasoning_mode": "off",
    "reasoning_level": "off",
    "reasoning_vendor_default": true,
    "score_mean_across_domains": 0.4810336213610023,
    "score_ci_lo": 0.43992676195057145,
    "score_ci_hi": 0.5229719244600196,
    "top100": 48,
    "n_cells": 126,
    "hallu_rate_mean_across_domains": 2.3015873015873014,
    "hallu_top100_across_domains": 42,
    "hallu_per_1k_tokens": 4.996812378310388,
    "contradicted_rate_mean": 0.7619047619047619,
    "extra_claims_rate_mean": 1.5396825396825398,
    "extra_false_rate_mean": 0.9722222222222222,
    "extra_unsupported_rate_mean": 0.5674603174603174,
    "score_mean_ex_family": 0.43721674416118855,
    "self_judge_delta": 0.04381687719981375,
    "n_cells_ex_family": 126,
    "false_claim_types": {
     "az": 0,
     "paragraph": 220,
     "study": 1,
     "number": 93,
     "other": 306
    },
    "n_failed": 0,
    "n_timeout": 0,
    "duration_median_s": 7.7,
    "duration_p90_s": 12.7,
    "reliability_source": "derived-from-raw"
   }
  ],
  "oracle_by_domain": [
   {
    "domain": "jura",
    "n_items": 35,
    "oracle_mean": 0.9536,
    "oracle_top100": 95,
    "best_single_model_id": "claude-opus-5",
    "best_single_model_mean": 0.9107,
    "best_single_model_top100": 91
   },
   {
    "domain": "medizin",
    "n_items": 30,
    "oracle_mean": 0.9604,
    "oracle_top100": 96,
    "best_single_model_id": "claude-opus-5",
    "best_single_model_mean": 0.92,
    "best_single_model_top100": 92
   },
   {
    "domain": "steuerrecht",
    "n_items": 31,
    "oracle_mean": 0.9175,
    "oracle_top100": 92,
    "best_single_model_id": "claude-opus-5",
    "best_single_model_mean": 0.8899,
    "best_single_model_top100": 89
   },
   {
    "domain": "wirtschaftsrecht",
    "n_items": 30,
    "oracle_mean": 0.8905,
    "oracle_top100": 89,
    "best_single_model_id": "claude-opus-5",
    "best_single_model_mean": 0.7989,
    "best_single_model_top100": 80
   }
  ],
  "cost": {
   "total_usd": 283.5037
  }
 }
}