{ "bench_version": 1, "track": "sequence", "model": "transformer_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.001, "temperature": 1.0 }, "sweep": [ { "cfg": { "lr": 0.001, "temperature": 0.7 }, "mean": 0.6538640260696411 }, { "cfg": { "lr": 0.001, "temperature": 1.0 }, "mean": 0.6523745656013489 }, { "cfg": { "lr": 0.003, "temperature": 0.7 }, "mean": 0.6653543412685394 }, { "cfg": { "lr": 0.003, "temperature": 1.0 }, "mean": 0.6606301814317703 }, { "cfg": { "lr": 0.006, "temperature": 0.7 }, "mean": 0.7325459271669388 }, { "cfg": { "lr": 0.006, "temperature": 1.0 }, "mean": 0.7683279514312744 } ], "full": { "mean": 0.6796555295586586, "std": 0.05923041132805598, "per_seed": [ 0.7301222085952759, 0.6001890897750854, 0.6041663885116577, 0.6750205755233765, 0.7771981358528137, 0.6423909664154053, 0.7304211258888245, 0.6777357459068298 ], "n": 8 } }, "idea": { "mean": 0.647096149623394, "std": 0.07594962062926218, "per_seed": [ 0.774990975856781, 0.5947428941726685, 0.5252495408058167, 0.6463086605072021, 0.7060304880142212, 0.5773214101791382, 0.7090907096862793, 0.6430345177650452 ], "n": 8 }, "comparison": { "delta_mean": -0.03255937993526459, "idea_wins": 7, "n_pairs": 8, "per_seed_diffs": [ 0.04486876726150513, -0.005446195602416992, -0.07891684770584106, -0.028711915016174316, -0.07116764783859253, -0.06506955623626709, -0.021330416202545166, -0.03470122814178467 ], "p_value": 0.0673, "mde": 0.033997422326021196, "mde_rel_pct": 5.002154892802501, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "trained_model": true, "M": 32, "observed_rmse_by_M": [ { "M": 8, "rmse": 0.4599074721336365 }, { "M": 16, "rmse": 0.35632288455963135 }, { "M": 32, "rmse": 0.22849832475185394 } ], "observed_slope": -0.5045800246521701, "predicted_slope": -0.5, "signed_rmse": 0.22849832475185394, "unsigned_rmse": 0.38780272006988525, "confirmed": true }, "idea_sweep": [ { "cfg": { "lr": 0.001, "m": 32 }, "result": { "mean": 0.7032602652907372, "std": 0.11299721951432162, "per_seed": [ 0.7925687432289124, 0.6065788269042969, 0.5842217206954956, 0.6156045198440552, 0.8893585801124573, 0.5813699960708618, 0.74463951587677, 0.8117402195930481 ], "n": 8 } }, { "cfg": { "lr": 0.003, "m": 32 }, "result": { "mean": 0.647096149623394, "std": 0.07594962062926218, "per_seed": [ 0.774990975856781, 0.5947428941726685, 0.5252495408058167, 0.6463086605072021, 0.7060304880142212, 0.5773214101791382, 0.7090907096862793, 0.6430345177650452 ], "n": 8 } }, { "cfg": { "lr": 0.006, "m": 32 }, "result": { "mean": 0.9036180377006531, "std": 0.0674238212671651, "per_seed": [ 0.9230982661247253, 0.871337890625, 0.8432034254074097, 0.7719256281852722, 0.9592460989952087, 0.9292699694633484, 1.0012027025222778, 0.9296603202819824 ], "n": 8 } } ], "sanity_check": { "rows": [ { "M": 16, "signed_rmse": 0.3650483780576059, "positive_rmse": 0.6758467186361339 }, { "M": 32, "signed_rmse": 0.20704154016476478, "positive_rmse": 0.6506880673538492 }, { "M": 64, "signed_rmse": 0.16207412164394386, "positive_rmse": 0.6397375154354732 }, { "M": 128, "signed_rmse": 0.12449882232224384, "positive_rmse": 0.6258620222659124 }, { "M": 256, "signed_rmse": 0.08409067839559449, "positive_rmse": 0.6281025092775255 } ], "observed_slope": -0.49699279208004493, "predicted_slope": -0.5, "positive_predicted_floor": 0.626260339536197, "passed": true }, "protocol": { "seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "n_train": 400, "n_test": 200, "epochs": 6, "baseline_grid": [ { "lr": 0.001, "temperature": 0.7 }, { "lr": 0.001, "temperature": 1.0 }, { "lr": 0.003, "temperature": 0.7 }, { "lr": 0.003, "temperature": 1.0 }, { "lr": 0.006, "temperature": 0.7 }, { "lr": 0.006, "temperature": 1.0 } ], "idea_grid": [ { "lr": 0.001, "m": 32 }, { "lr": 0.003, "m": 32 }, { "lr": 0.006, "m": 32 } ] } }