| { | |
| "splits": [ | |
| { | |
| "leakage_audit": { | |
| "passed": true, | |
| "train_records": 672, | |
| "test_records": 327, | |
| "compound_overlap": [], | |
| "formula_overlap": [] | |
| }, | |
| "n_test": 327, | |
| "baseline_accuracy": 0.7889908256880734, | |
| "learned_accuracy": 0.6972477064220184, | |
| "paired_advantage": { | |
| "delta": -0.09174311926605505, | |
| "ci95_low": -0.13761467889908258, | |
| "ci95_high": -0.045871559633027525 | |
| }, | |
| "name": "seed_1" | |
| }, | |
| { | |
| "leakage_audit": { | |
| "passed": true, | |
| "train_records": 623, | |
| "test_records": 309, | |
| "compound_overlap": [], | |
| "formula_overlap": [] | |
| }, | |
| "n_test": 309, | |
| "baseline_accuracy": 0.8414239482200647, | |
| "learned_accuracy": 0.6893203883495146, | |
| "paired_advantage": { | |
| "delta": -0.15210355987055016, | |
| "ci95_low": -0.20388349514563106, | |
| "ci95_high": -0.10355987055016182 | |
| }, | |
| "name": "seed_2" | |
| }, | |
| { | |
| "leakage_audit": { | |
| "passed": true, | |
| "train_records": 625, | |
| "test_records": 340, | |
| "compound_overlap": [], | |
| "formula_overlap": [] | |
| }, | |
| "n_test": 340, | |
| "baseline_accuracy": 0.8088235294117647, | |
| "learned_accuracy": 0.7058823529411765, | |
| "paired_advantage": { | |
| "delta": -0.10294117647058823, | |
| "ci95_low": -0.15, | |
| "ci95_high": -0.05588235294117647 | |
| }, | |
| "name": "seed_3" | |
| } | |
| ], | |
| "verdict": { | |
| "decision": "negative_result", | |
| "reason": "learned representation did not reliably beat composition baseline", | |
| "valid_splits": 3, | |
| "minimum_splits": 3, | |
| "required_margin": 0.0 | |
| } | |
| } | |