| { |
| "protocol_version": 1, |
| "status": "embeddings_exported_benchmark_not_run", |
| "dataset": { |
| "path": "data/empirical_dataset_v9_plus_wisemoor_clean.jsonl", |
| "sha256": "9219f4adf6a76e9e494bfbe9b0be80e3ef092f2e395ae477fdf8977df1ae2cd6", |
| "records": 5708 |
| }, |
| "checkpoint": { |
| "path": "models/pimt_clean_rerun/pimt_clean_rerun.pt", |
| "sha256": "e7c1794fd1ed8f67e1ee910af45cdd4f5d8097eb014cbda3138a8f10cd939d99" |
| }, |
| "git_commit": "3c1e178cde8eaebf9dcc14c2e94001fcc88f87e7", |
| "split": { |
| "method": "active-molecule assignment; mixed-boundary formulas excluded", |
| "train_ratio": 0.6, |
| "seeds": [ |
| 1, |
| 2, |
| 3 |
| ], |
| "minimum_records_per_genre_per_partition": 5 |
| }, |
| "benchmark": { |
| "probe": "training-standardized nearest centroid", |
| "bootstrap_samples": 10000, |
| "bootstrap_seed": 8675309, |
| "minimum_valid_splits": 3, |
| "required_margin": 0.0, |
| "decision": "support iff every valid split has paired CI95 lower bound > required_margin" |
| }, |
| "manifest": "artifacts/genre_eval_frozen/genre_split_manifest.json", |
| "coverage_audit": [ |
| { |
| "name": "seed_1", |
| "seed": 1, |
| "train_records": 672, |
| "test_records": 327, |
| "excluded_records": 4709, |
| "train_genres": { |
| "amber_oriental": 67, |
| "citrus_cologne": 70, |
| "floral_woody": 65, |
| "fougere": 53, |
| "wildcard": 417 |
| }, |
| "test_genres": { |
| "amber_oriental": 15, |
| "citrus_cologne": 16, |
| "floral_woody": 13, |
| "fougere": 10, |
| "wildcard": 273 |
| }, |
| "minimum_per_genre": 5, |
| "coverage_passed": true |
| }, |
| { |
| "name": "seed_2", |
| "seed": 2, |
| "train_records": 623, |
| "test_records": 309, |
| "excluded_records": 4776, |
| "train_genres": { |
| "amber_oriental": 51, |
| "citrus_cologne": 65, |
| "floral_woody": 53, |
| "fougere": 43, |
| "wildcard": 411 |
| }, |
| "test_genres": { |
| "amber_oriental": 14, |
| "citrus_cologne": 11, |
| "floral_woody": 22, |
| "fougere": 9, |
| "wildcard": 253 |
| }, |
| "minimum_per_genre": 5, |
| "coverage_passed": true |
| }, |
| { |
| "name": "seed_3", |
| "seed": 3, |
| "train_records": 625, |
| "test_records": 340, |
| "excluded_records": 4743, |
| "train_genres": { |
| "amber_oriental": 75, |
| "citrus_cologne": 54, |
| "floral_woody": 51, |
| "fougere": 47, |
| "wildcard": 398 |
| }, |
| "test_genres": { |
| "amber_oriental": 16, |
| "citrus_cologne": 14, |
| "floral_woody": 21, |
| "fougere": 15, |
| "wildcard": 274 |
| }, |
| "minimum_per_genre": 5, |
| "coverage_passed": true |
| } |
| ], |
| "embedding_export": { |
| "pooling": "unmasked mean over time and active ingredient tokens", |
| "dimension": 256, |
| "unique_records": 1511 |
| } |
| } |
|
|