Troubadour / RESULTS.json
noahmclaughlin's picture
Release nine audited splits and completed Troubadour-Bench results
ed8341f verified
Raw History Blame Contribute Delete
30.4 kB
{
"research_question": "Can post-training teach a model street smarts so that it can learn and apply unfamiliar social rules?",
"selection": {
"frozen_at": "2026-11-17T08:58:58.379230+00:00",
"base": "mlx-community/Qwen3-4B-Instruct-2507-4bit",
"base_revision": "50d427756c6b1b2fe0c0a10f67fbda1fc8e82c1b",
"recipe": "choice-slow-seed17",
"seed": 17,
"selection_code_revision": "482939964db176068df106d2ae66924703cb8055",
"checkpoint_microsteps": 1200,
"source_checkpoint": "artifacts/slow-probe-1200/adapters.safetensors",
"adapter_sha256": "875c12bbecfed33888290b438a43a3f63c5f076bd0d16528018289dabdf3976c",
"adapter_config_sha256": "55c72bc1943e558708b93a987669b21002462ca0ce68fbcf250feb58f3a3b57e",
"training_config": "configs/choice-slow-seed17.yaml",
"training_config_sha256": "4e6831c4d753af179b7048823c3debb0474850e24d032262c316b647b567b0a8",
"training_data": "data/choice/train.jsonl",
"training_data_sha256": "f21e7cf04ed8da7a46ccf3aad980b61fa917661195ac08a20e24f38d467ed22d",
"split_sha256": "f07e2072e199729f4142c4221722d41cfff76b8689d677b1500a2341e77616d6",
"criteria_sha256": "ba6c3804c4dd9b66975824613d38d8fa36436712b5ff18be57f13fd8182b562b",
"base_calibration_sha256": "d3a4af0299bbbbd6856a846340accc38f6a7ed74b8c14a166bfe3979f43281b3",
"candidate_calibration_sha256": "0f64b70fd4fb3c057ee85a00b36248e907d5d0fffd39f7f0c9284ae9cae10944",
"development_predictions_sha256": "af73fd25cba94d36f17ba889c252577b497b7c747afaeccc01d9ffdc6c5ed019",
"renderer_sha256": "20488e4138d2c0ba7c895e707506dc7baa1887a274cf118d045c577dd1295b80",
"scoring_protocol": "constrained-choice-v4-reference",
"runtime_versions": {
"mlx": "0.32.4",
"mlx-lm": "0.32.2",
"transformers": "5.23.0"
},
"selection_reason": "None of the eight bounded v4 candidates met the mandatory IID development criteria. Under the recorded fallback rule, select the highest joint accuracy, then uncertainty recall, then earlier microstep. The long-run 1200 checkpoint tied balanced-rate 600 at 48.75% accuracy and won on uncertainty recall (21.25% versus 11.25%). Publish as an exploratory research checkpoint; do not claim validated improvement from development scores. No final encounters were generated or inspected.",
"final_data_seen": false,
"secondary_diagnostics": {
"learning_curve_first_n": 60,
"learning_curve_ks": [
0,
1,
2,
4,
8
],
"robustness_first_n": 48,
"robustness_styles": [
"prose",
"symbols"
],
"scope": "Fixed prefixes of the reserved episode order; exploratory small-cohort diagnostics"
}
},
"criteria": {
"primary": "constrained-choice A/B/U accuracy at 8 observations on reserved IID society pairs",
"minimum_raw_qwen_gain": 0.1,
"paired_cluster_bootstrap_lower_bound_must_exceed": 0,
"minimum_answerable_gain": 0.1,
"maximum_uncertainty_recall_drop": 0.1,
"minimum_gain_over_dev_calibrated_qwen": 0.05,
"transfer_gain_must_be_positive": true,
"replicate_seeds": [
17,
29,
43
],
"published_seed_preselected": 17,
"heuristic_comparison": "Report explicitly; do not claim best method if the nearest-example heuristic is stronger",
"status": "Frozen after base pilot, before candidate development results or final generation"
},
"checks": {
"raw_qwen_gain": false,
"positive_cluster_interval": false,
"answerable_gain": true,
"uncertainty_recall_preserved": false,
"gain_over_calibrated_qwen": false,
"positive_transfer_gain": true
},
"passes_primary_criteria": false,
"primary": {
"base": {
"n": 306,
"accuracy": 0.4150326797385621,
"balanced_accuracy": 0.4150326797385621,
"answerable_accuracy": 0.3333333333333333,
"conditional_AB_accuracy": 0.6470588235294118,
"uncertainty_ranking_auc": 0.48776912725874666,
"uncertainty_recall": 0.5784313725490197,
"uncertainty_precision": 0.35542168674698793,
"paired_both_correct": 0.2875816993464052,
"paired_answerable_both_correct": 0.17647058823529413,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 64,
"A": 76,
"U": 166
},
"seconds": 823.5870471000671
},
"candidate": {
"n": 306,
"accuracy": 0.4542483660130719,
"balanced_accuracy": 0.4542483660130719,
"answerable_accuracy": 0.6274509803921569,
"conditional_AB_accuracy": 0.7450980392156863,
"uncertainty_ranking_auc": 0.46145713187235676,
"uncertainty_recall": 0.10784313725490197,
"uncertainty_precision": 0.23404255319148937,
"paired_both_correct": 0.3464052287581699,
"paired_answerable_both_correct": 0.5,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 144,
"A": 115,
"U": 47
},
"seconds": 261.5247766971588
},
"society_pairs": 51,
"cluster_mean_gain": 0.039215686274509796,
"cluster_bootstrap_95ci": [
-0.013071895424836605,
0.09477124183006536
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"against_dev_calibrated_qwen": {
"base": {
"n": 306,
"accuracy": 0.41830065359477125,
"balanced_accuracy": 0.41830065359477125,
"answerable_accuracy": 0.6176470588235294,
"conditional_AB_accuracy": 0.6470588235294118,
"uncertainty_ranking_auc": 0.48776912725874666,
"uncertainty_recall": 0.0196078431372549,
"uncertainty_precision": 0.16666666666666666,
"paired_both_correct": 0.28104575163398693,
"paired_answerable_both_correct": 0.4215686274509804,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 138,
"A": 156,
"U": 12
},
"seconds": 823.5870471000671
},
"candidate": {
"n": 306,
"accuracy": 0.4542483660130719,
"balanced_accuracy": 0.4542483660130719,
"answerable_accuracy": 0.6274509803921569,
"conditional_AB_accuracy": 0.7450980392156863,
"uncertainty_ranking_auc": 0.46145713187235676,
"uncertainty_recall": 0.10784313725490197,
"uncertainty_precision": 0.23404255319148937,
"paired_both_correct": 0.3464052287581699,
"paired_answerable_both_correct": 0.5,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 144,
"A": 115,
"U": 47
},
"seconds": 261.5247766971588
},
"society_pairs": 51,
"cluster_mean_gain": 0.03594771241830065,
"cluster_bootstrap_95ci": [
-0.009803921568627449,
0.08169934640522875
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"transfer": {
"base": {
"n": 150,
"accuracy": 0.36666666666666664,
"balanced_accuracy": 0.3666666666666667,
"answerable_accuracy": 0.26,
"conditional_AB_accuracy": 0.55,
"uncertainty_ranking_auc": 0.5031,
"uncertainty_recall": 0.58,
"uncertainty_precision": 0.3333333333333333,
"paired_both_correct": 0.18666666666666668,
"paired_answerable_both_correct": 0.08,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 87,
"A": 34,
"B": 29
},
"seconds": 405.3112745285034
},
"candidate": {
"n": 150,
"accuracy": 0.4066666666666667,
"balanced_accuracy": 0.4066666666666667,
"answerable_accuracy": 0.48,
"conditional_AB_accuracy": 0.6,
"uncertainty_ranking_auc": 0.4787,
"uncertainty_recall": 0.26,
"uncertainty_precision": 0.40625,
"paired_both_correct": 0.30666666666666664,
"paired_answerable_both_correct": 0.34,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 32,
"B": 68,
"A": 50
},
"seconds": 128.30947995185852
},
"society_pairs": 25,
"cluster_mean_gain": 0.039999999999999994,
"cluster_bootstrap_95ci": [
-0.05333333333333333,
0.13333333333333333
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"opposite_query_stress": {
"base": {
"n": 204,
"accuracy": 0.28431372549019607,
"balanced_accuracy": 0.28431372549019607,
"answerable_accuracy": 0.28431372549019607,
"conditional_AB_accuracy": 0.5931372549019608,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.10784313725490197,
"paired_answerable_both_correct": 0.10784313725490197,
"contrast_both_queries_correct": 0.058823529411764705,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 118,
"A": 50,
"B": 36
},
"seconds": 554.2962846755981
},
"candidate": {
"n": 204,
"accuracy": 0.5343137254901961,
"balanced_accuracy": 0.5343137254901961,
"answerable_accuracy": 0.5343137254901961,
"conditional_AB_accuracy": 0.5980392156862745,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.39215686274509803,
"paired_answerable_both_correct": 0.39215686274509803,
"contrast_both_queries_correct": 0.13725490196078433,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 93,
"A": 81,
"U": 30
},
"seconds": 175.2643644809723
},
"society_pairs": 51,
"cluster_mean_gain": 0.25,
"cluster_bootstrap_95ci": [
0.19117647058823528,
0.3088235294117647
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"transfer_against_dev_calibrated_qwen": {
"base": {
"n": 150,
"accuracy": 0.38666666666666666,
"balanced_accuracy": 0.3866666666666667,
"answerable_accuracy": 0.54,
"conditional_AB_accuracy": 0.55,
"uncertainty_ranking_auc": 0.5031,
"uncertainty_recall": 0.08,
"uncertainty_precision": 0.4,
"paired_both_correct": 0.22666666666666666,
"paired_answerable_both_correct": 0.32,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"A": 71,
"B": 69,
"U": 10
},
"seconds": 405.3112745285034
},
"candidate": {
"n": 150,
"accuracy": 0.4066666666666667,
"balanced_accuracy": 0.4066666666666667,
"answerable_accuracy": 0.48,
"conditional_AB_accuracy": 0.6,
"uncertainty_ranking_auc": 0.4787,
"uncertainty_recall": 0.26,
"uncertainty_precision": 0.40625,
"paired_both_correct": 0.30666666666666664,
"paired_answerable_both_correct": 0.34,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 32,
"B": 68,
"A": 50
},
"seconds": 128.30947995185852
},
"society_pairs": 25,
"cluster_mean_gain": 0.019999999999999997,
"cluster_bootstrap_95ci": [
-0.05333333333333333,
0.09333333333333332
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"contrast_against_dev_calibrated_qwen": {
"base": {
"n": 204,
"accuracy": 0.5882352941176471,
"balanced_accuracy": 0.588235294117647,
"answerable_accuracy": 0.5882352941176471,
"conditional_AB_accuracy": 0.5931372549019608,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.39215686274509803,
"paired_answerable_both_correct": 0.39215686274509803,
"contrast_both_queries_correct": 0.2647058823529412,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 86,
"A": 117,
"U": 1
},
"seconds": 554.2962846755981
},
"candidate": {
"n": 204,
"accuracy": 0.5343137254901961,
"balanced_accuracy": 0.5343137254901961,
"answerable_accuracy": 0.5343137254901961,
"conditional_AB_accuracy": 0.5980392156862745,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.39215686274509803,
"paired_answerable_both_correct": 0.39215686274509803,
"contrast_both_queries_correct": 0.13725490196078433,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 93,
"A": 81,
"U": 30
},
"seconds": 175.2643644809723
},
"society_pairs": 51,
"cluster_mean_gain": -0.05392156862745098,
"cluster_bootstrap_95ci": [
-0.11274509803921569,
0.004901960784313725
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"secondary_both_dev_calibrated": {
"primary": {
"base": {
"n": 306,
"accuracy": 0.41830065359477125,
"balanced_accuracy": 0.41830065359477125,
"answerable_accuracy": 0.6176470588235294,
"conditional_AB_accuracy": 0.6470588235294118,
"uncertainty_ranking_auc": 0.48776912725874666,
"uncertainty_recall": 0.0196078431372549,
"uncertainty_precision": 0.16666666666666666,
"paired_both_correct": 0.28104575163398693,
"paired_answerable_both_correct": 0.4215686274509804,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 138,
"A": 156,
"U": 12
},
"seconds": 823.5870471000671
},
"candidate": {
"n": 306,
"accuracy": 0.4542483660130719,
"balanced_accuracy": 0.4542483660130719,
"answerable_accuracy": 0.6274509803921569,
"conditional_AB_accuracy": 0.7450980392156863,
"uncertainty_ranking_auc": 0.46145713187235676,
"uncertainty_recall": 0.10784313725490197,
"uncertainty_precision": 0.23404255319148937,
"paired_both_correct": 0.3464052287581699,
"paired_answerable_both_correct": 0.5,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 144,
"A": 115,
"U": 47
},
"seconds": 261.5247766971588
},
"society_pairs": 51,
"cluster_mean_gain": 0.03594771241830065,
"cluster_bootstrap_95ci": [
-0.009803921568627449,
0.08169934640522875
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"transfer": {
"base": {
"n": 150,
"accuracy": 0.38666666666666666,
"balanced_accuracy": 0.3866666666666667,
"answerable_accuracy": 0.54,
"conditional_AB_accuracy": 0.55,
"uncertainty_ranking_auc": 0.5031,
"uncertainty_recall": 0.08,
"uncertainty_precision": 0.4,
"paired_both_correct": 0.22666666666666666,
"paired_answerable_both_correct": 0.32,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"A": 71,
"B": 69,
"U": 10
},
"seconds": 405.3112745285034
},
"candidate": {
"n": 150,
"accuracy": 0.4066666666666667,
"balanced_accuracy": 0.4066666666666667,
"answerable_accuracy": 0.48,
"conditional_AB_accuracy": 0.6,
"uncertainty_ranking_auc": 0.4787,
"uncertainty_recall": 0.26,
"uncertainty_precision": 0.40625,
"paired_both_correct": 0.30666666666666664,
"paired_answerable_both_correct": 0.34,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 32,
"B": 68,
"A": 50
},
"seconds": 128.30947995185852
},
"society_pairs": 25,
"cluster_mean_gain": 0.019999999999999997,
"cluster_bootstrap_95ci": [
-0.05333333333333333,
0.09333333333333332
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"contrast": {
"base": {
"n": 204,
"accuracy": 0.5882352941176471,
"balanced_accuracy": 0.588235294117647,
"answerable_accuracy": 0.5882352941176471,
"conditional_AB_accuracy": 0.5931372549019608,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.39215686274509803,
"paired_answerable_both_correct": 0.39215686274509803,
"contrast_both_queries_correct": 0.2647058823529412,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 86,
"A": 117,
"U": 1
},
"seconds": 554.2962846755981
},
"candidate": {
"n": 204,
"accuracy": 0.5343137254901961,
"balanced_accuracy": 0.5343137254901961,
"answerable_accuracy": 0.5343137254901961,
"conditional_AB_accuracy": 0.5980392156862745,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.39215686274509803,
"paired_answerable_both_correct": 0.39215686274509803,
"contrast_both_queries_correct": 0.13725490196078433,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 93,
"A": 81,
"U": 30
},
"seconds": 175.2643644809723
},
"society_pairs": 51,
"cluster_mean_gain": -0.05392156862745098,
"cluster_bootstrap_95ci": [
-0.11274509803921569,
0.004901960784313725
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
}
},
"calibration_biases": {
"base": -4.0,
"candidate": 0.0,
"fit_on": "development only; primary model remains unadjusted"
},
"nonparametric_baselines": {
"primary": {
"nearest_example": {
"n": 306,
"accuracy": 0.5686274509803921,
"balanced_accuracy": 0.5686274509803921,
"answerable_accuracy": 0.7647058823529411,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 0.17647058823529413,
"uncertainty_precision": 0.28125,
"paired_both_correct": 0.5686274509803921,
"paired_answerable_both_correct": 0.7647058823529411,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 64,
"A": 121,
"B": 121
},
"seconds": 0
},
"majority_label": {
"n": 306,
"accuracy": 0.45751633986928103,
"balanced_accuracy": 0.45751633986928103,
"answerable_accuracy": 0.6274509803921569,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 0.11764705882352941,
"uncertainty_precision": 0.23076923076923078,
"paired_both_correct": 0.45751633986928103,
"paired_answerable_both_correct": 0.6274509803921569,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 127,
"A": 127,
"U": 52
},
"seconds": 0
},
"always_U": {
"n": 306,
"accuracy": 0.3333333333333333,
"balanced_accuracy": 0.3333333333333333,
"answerable_accuracy": 0.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 1.0,
"uncertainty_precision": 0.3333333333333333,
"paired_both_correct": 0.3333333333333333,
"paired_answerable_both_correct": 0.0,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 306
},
"seconds": 0
},
"reference_solver": {
"n": 306,
"accuracy": 1.0,
"balanced_accuracy": 1.0,
"answerable_accuracy": 1.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 1.0,
"uncertainty_precision": 1.0,
"paired_both_correct": 1.0,
"paired_answerable_both_correct": 1.0,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 102,
"A": 102,
"B": 102
},
"seconds": 0
}
},
"transfer": {
"nearest_example": {
"n": 150,
"accuracy": 0.52,
"balanced_accuracy": 0.52,
"answerable_accuracy": 0.58,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 0.4,
"uncertainty_precision": 0.37037037037037035,
"paired_both_correct": 0.52,
"paired_answerable_both_correct": 0.58,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"B": 48,
"A": 48,
"U": 54
},
"seconds": 0
},
"majority_label": {
"n": 150,
"accuracy": 0.3333333333333333,
"balanced_accuracy": 0.3333333333333333,
"answerable_accuracy": 0.42,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 0.16,
"uncertainty_precision": 0.25,
"paired_both_correct": 0.3333333333333333,
"paired_answerable_both_correct": 0.42,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"A": 59,
"B": 59,
"U": 32
},
"seconds": 0
},
"always_U": {
"n": 150,
"accuracy": 0.3333333333333333,
"balanced_accuracy": 0.3333333333333333,
"answerable_accuracy": 0.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 1.0,
"uncertainty_precision": 0.3333333333333333,
"paired_both_correct": 0.3333333333333333,
"paired_answerable_both_correct": 0.0,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 150
},
"seconds": 0
},
"reference_solver": {
"n": 150,
"accuracy": 1.0,
"balanced_accuracy": 1.0,
"answerable_accuracy": 1.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 1.0,
"uncertainty_precision": 1.0,
"paired_both_correct": 1.0,
"paired_answerable_both_correct": 1.0,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 50,
"A": 50,
"B": 50
},
"prediction_counts": {
"U": 50,
"A": 50,
"B": 50
},
"seconds": 0
}
},
"contrast": {
"nearest_example": {
"n": 204,
"accuracy": 0.7058823529411765,
"balanced_accuracy": 0.7058823529411765,
"answerable_accuracy": 0.7058823529411765,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.7058823529411765,
"paired_answerable_both_correct": 0.7058823529411765,
"contrast_both_queries_correct": 0.49019607843137253,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 54,
"B": 75,
"A": 75
},
"seconds": 0
},
"majority_label": {
"n": 204,
"accuracy": 0.4215686274509804,
"balanced_accuracy": 0.4215686274509804,
"answerable_accuracy": 0.4215686274509804,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.4215686274509804,
"paired_answerable_both_correct": 0.4215686274509804,
"contrast_both_queries_correct": 0.0,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 86,
"A": 86,
"U": 32
},
"seconds": 0
},
"always_U": {
"n": 204,
"accuracy": 0.0,
"balanced_accuracy": 0.0,
"answerable_accuracy": 0.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": 0.0,
"paired_both_correct": 0.0,
"paired_answerable_both_correct": 0.0,
"contrast_both_queries_correct": 0.0,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 204
},
"seconds": 0
},
"reference_solver": {
"n": 204,
"accuracy": 1.0,
"balanced_accuracy": 1.0,
"answerable_accuracy": 1.0,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": null,
"uncertainty_precision": null,
"paired_both_correct": 1.0,
"paired_answerable_both_correct": 1.0,
"contrast_both_queries_correct": 1.0,
"invalid": 0,
"expected_counts": {
"A": 102,
"B": 102
},
"prediction_counts": {
"A": 102,
"B": 102
},
"seconds": 0
}
}
},
"candidate_vs_nearest_example": {
"base": {
"n": 306,
"accuracy": 0.5686274509803921,
"balanced_accuracy": 0.5686274509803921,
"answerable_accuracy": 0.7647058823529411,
"conditional_AB_accuracy": null,
"uncertainty_ranking_auc": null,
"uncertainty_recall": 0.17647058823529413,
"uncertainty_precision": 0.28125,
"paired_both_correct": 0.5686274509803921,
"paired_answerable_both_correct": 0.7647058823529411,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"U": 64,
"A": 121,
"B": 121
},
"seconds": 0
},
"candidate": {
"n": 306,
"accuracy": 0.4542483660130719,
"balanced_accuracy": 0.4542483660130719,
"answerable_accuracy": 0.6274509803921569,
"conditional_AB_accuracy": 0.7450980392156863,
"uncertainty_ranking_auc": 0.46145713187235676,
"uncertainty_recall": 0.10784313725490197,
"uncertainty_precision": 0.23404255319148937,
"paired_both_correct": 0.3464052287581699,
"paired_answerable_both_correct": 0.5,
"contrast_both_queries_correct": null,
"invalid": 0,
"expected_counts": {
"U": 102,
"A": 102,
"B": 102
},
"prediction_counts": {
"B": 144,
"A": 115,
"U": 47
},
"seconds": 261.5247766971588
},
"society_pairs": 51,
"cluster_mean_gain": -0.11437908496732026,
"cluster_bootstrap_95ci": [
-0.1895424836601307,
-0.0392156862745098
],
"method": "paired bootstrap over complement-society clusters, equal cluster weighting; 10000 replicates"
},
"interpretation": "Synthetic finite-grammar rule induction in social language; no claim of real-world social competence. Independent-seed and robustness results must be reported separately."
}