{ "completion": "complete", "coverage": { "denominator": 184, "scored": 184, "unscored": 0 }, "coverage_context": { "not_selected": 0, "selected": 184, "unscored_file_scope": "selected inputs only; not_selected counted separately" }, "created_at": "2026-09-21T12:05:00.053763+00:00", "dataset_id": "flip2-rhomax-by-wild-type", "environment": { "architecture": "arm64", "container_digest": "unreported", "container_runtime": "unreported", "platform": "macOS-26.6.2-arm64-arm-64bit", "python": "3.11.13", "rewirebench": "0.4.0", "sdk_code_sha256": "08cbbaac2aa8f60af5c43af7a83410e9690b44c5e4d162fbe1e5766885952fce", "sif_sha256": "unreported" }, "execution": { "adapter": "ESM2Embeddings", "adapter_provenance": { "checkpoint_sha256": "7f21e80e61d16a71735163ef555d3009afb0c98da74c48e29df08606973cc55e", "checkpoint_url": "https://dl.fbaipublicfiles.com/fair-esm/models/esm2_t12_35M_UR50D.pt", "device": "cpu", "fair_esm_version": "2.0.0", "implementation_sha256": "bb46676af95050012cf0bf8a8309581b400f660d2722519889693f4084559d73", "maximum_sequence_length": 1022, "model": "esm2_t12_35M_UR50D", "published_result_reproduction": false, "reference": "https://github.com/facebookresearch/esm/blob/2b369911bb5b4b0dda914521b9475cad1656b2ac/README.md#usage", "representation_layer": 12, "strategy": "frozen_final_layer_residue_mean_excluding_special_tokens", "training_overlap": "unreported; UniRef50 pretraining may overlap benchmark proteins" }, "batch_size": 4, "fitting": "protocol_owned_probe", "inference_and_fit_seconds": 64.14407666699844, "mode": "local_adapter", "prediction_type": "embedding" }, "independently_reproduced": false, "input_information": "Complete amino-acid sequences; no labels enter encoder; no MSA/templates", "kind": "rewire_local_evaluation", "metrics": { "n": 184, "ndcg": 0.9072580204736261, "spearman": -0.2217595033467806 }, "model": { "name": "ESM-2 35M frozen residue-mean embeddings + fixed ridge (Rewire)", "training_overlap": "UniRef50 pretraining overlap is unreported; the linear head fits only archived training rows" }, "model_configuration": { "checkpoint": "esm2_t12_35M_UR50D", "device": "cpu", "dimension": 480, "encoder_frozen": true, "head": { "alpha": 10.0, "feature_scaling": "none", "fit_intercept": true, "max_iter": 1000000, "name": "rewire-frozen-embedding-ridge-v1", "refit": "train only", "relationship_to_published_baseline": "Frozen-embedding extension of upstream alpha-10 ridge. The published baseline uses one-hot features and scales targets using train plus validation rows. This extension is not that published baseline.", "solver": "auto", "target_scaling": "StandardScaler fitted only on train labels", "tol": 1e-05, "validation_use": "none; fixed hyperparameters" }, "pooling": "mean over residues excluding BOS/EOS and padding", "representation_layer": 12, "scope": "one complete Rhomax split, not FLIP2 suite", "torch_seed": 0, "torch_threads": 4, "validation_used": false }, "predictions_sha256": "96e8a0c6663dd0a471b984fb86506e483d252a6036a9617cf5f4487664274abd", "prepared_sha256": "f417d43497e0c8c453bd7c66cfea05959b4d6d47462283096e42a2dae8b72419", "protocol_configuration": { "adapter_input_contract": "opaque-sequence-inputs-v1", "canonical_split_counts": { "test": 184, "train": 584, "validation": 116 }, "canonical_test_count": 184, "data_verification": "pinned_source_bytes", "dataset": "rhomax", "embedding_probe": { "alpha": 10.0, "feature_scaling": "none", "fit_intercept": true, "max_iter": 1000000, "name": "rewire-frozen-embedding-ridge-v1", "refit": "train only", "relationship_to_published_baseline": "Frozen-embedding extension of upstream alpha-10 ridge. The published baseline uses one-hot features and scales targets using train plus validation rows. This extension is not that published baseline.", "solver": "auto", "target_scaling": "StandardScaler fitted only on train labels", "tol": 1e-05, "validation_use": "none; fixed hyperparameters" }, "input_representation": "amino-acid sequence", "metric": "spearman", "metric_direction": "higher", "metrics": [ "spearman", "ndcg" ], "prepared_rows_sha256": "d86fe3926529df72649f4638ffa0d89ec62460d831ceb62ad47ff86249469a12", "smoke_limit_per_split": null, "source_note": null, "source_reuse_terms": "CC-BY-4.0 archive; source attribution under resources/flip2", "split": "by_wild_type", "suite_complete": false, "target_units": "nm", "task": "regression" }, "protocol_id": "flip2-fitness-v1", "protocol_results": { "complete": true, "coverage": { "denominator": 184, "scored": 184, "selected": 184, "unscored": 0 }, "metrics": { "n": 184, "ndcg": 0.9072580204736261, "spearman": -0.2217595033467806 }, "metrics_scope": "Scored test rows of this archived split only; no suite aggregate", "protocol_id": "flip2-fitness-v1", "protocol_notes": [], "scope": "full", "score_direction": "higher", "suite_complete": false, "unavailable_metrics": {}, "unscored_reasons": { "missing_prediction": 0, "not_selected": 0 } }, "protocol_version": "1", "provenance": { "data_verification": "pinned_source_bytes", "dataset_version": "zenodo-18433203-v3", "evaluator_url": "https://github.com/J-SNACKKB/FLIP/blob/62cace8735f5610e2743cf06ce0f944b37fffaa6/baselines/aggregate.py#L131-L135", "source_csv_sha256": "8e78f6a16cd5298131dca83130d4b94cc0ab4ae9c699460880441c19a65f656f", "source_sha256": "7c6d2f02cb89310378ac9897c321fbbd909fb0757ac88803dcce84f5f6b05ce3", "source_url": "https://zenodo.org/api/records/18433203/files/rhomax/by_wild_type.csv.gz/content", "split_origin": "Archived assignments: validation=True held out of training", "upstream_revision": "62cace8735f5610e2743cf06ce0f944b37fffaa6" }, "review_status": "unreviewed_local_run", "schema_version": "1.0", "scope": "full", "timing_seconds": { "evaluation": 0.1622792090056464 } }