{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "gate_mid": 0.08, "gate_slope": 3.0 }, "sweep": [ { "cfg": { "lr": 0.001, "gate_mid": 0.08, "gate_slope": 3.0 }, "mean": 0.001094405582989566 }, { "cfg": { "lr": 0.003, "gate_mid": 0.08, "gate_slope": 3.0 }, "mean": 0.0005730643169954419 }, { "cfg": { "lr": 0.01, "gate_mid": 0.08, "gate_slope": 3.0 }, "mean": 0.0001522819948149845 } ], "full": { "mean": 0.00012016432719974546, "std": 5.673950443795311e-05, "per_seed": [ 0.00023659199359826744, 0.000133594949147664, 7.917142647784203e-05, 0.00015976961003616452, 6.463151657953858e-05, 9.197137842420489e-05, 5.368743586586788e-05, 0.0001418963074684143 ], "n": 8 } }, "idea": { "mean": 4.0656192368260236e-06, "std": 1.5779659647239028e-07, "per_seed": [ 4.112095211894484e-06, 3.850047960440861e-06, 3.88751095670159e-06, 4.318793344282312e-06, 3.984665454481728e-06, 3.9648107303946745e-06, 4.202797754260246e-06, 4.2042324821522925e-06 ], "n": 8 }, "comparison": { "delta_mean": -0.00011609870796291943, "idea_wins": 8, "n_pairs": 8, "per_seed_diffs": [ -0.00023247989838637295, -0.00012974490118722315, -7.528391552114044e-05, -0.0001554508166918822, -6.0646851125056855e-05, -8.800656769381021e-05, -4.948463811160764e-05, -0.00013769207498626201 ], "p_value": 0.0081, "mde": 5.06773245421868e-05, "mde_rel_pct": 42.17335187833861, "verdict": "idea better (significant)", "system_worked": true }, "mechanism_signature": { "trained_model": true, "defect_low_closure": 0.0020928073208779097, "defect_high_closure": 0.01778038963675499, "defect_closure_corr": 0.9902236612013674, "background_rmse": 0.008130352012813091, "prediction": "closure should rise with defect and fall as background becomes accurate", "confirmed": true }, "idea_sweep": [ { "cfg": { "lr": 0.001, "gate_mid": 0.08, "gate_slope": 3.0 }, "result": { "mean": 4.137774908485881e-06, "std": 1.5718023689435836e-07, "per_seed": [ 4.205145160085522e-06, 3.896001544489991e-06, 4.057551450387109e-06, 4.373848241812084e-06, 4.077617177244974e-06, 3.985380772064673e-06, 4.350635663286084e-06, 4.156019258516608e-06 ], "n": 8 } }, { "cfg": { "lr": 0.003, "gate_mid": 0.08, "gate_slope": 3.0 }, "result": { "mean": 4.0656192368260236e-06, "std": 1.5779659647239028e-07, "per_seed": [ 4.112095211894484e-06, 3.850047960440861e-06, 3.88751095670159e-06, 4.318793344282312e-06, 3.984665454481728e-06, 3.9648107303946745e-06, 4.202797754260246e-06, 4.2042324821522925e-06 ], "n": 8 } }, { "cfg": { "lr": 0.01, "gate_mid": 0.08, "gate_slope": 3.0 }, "result": { "mean": 4.102830018837267e-06, "std": 1.5946786154305604e-07, "per_seed": [ 4.118781362194568e-06, 3.881638804159593e-06, 3.890218522428768e-06, 4.2627821130736265e-06, 4.292129688110435e-06, 3.9507722249254584e-06, 4.203915068501374e-06, 4.222402367304312e-06 ], "n": 8 } } ], "protocol": { "paired_seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "sweep_seeds": [ 0, 1, 2, 3 ], "epochs": 15, "n_train": 1000, "n_test": 400, "same_architecture": true, "baseline_grid": [ { "lr": 0.001, "gate_mid": 0.08, "gate_slope": 3.0 }, { "lr": 0.003, "gate_mid": 0.08, "gate_slope": 3.0 }, { "lr": 0.01, "gate_mid": 0.08, "gate_slope": 3.0 } ], "idea_grid": [ { "lr": 0.001, "gate_mid": 0.08, "gate_slope": 3.0 }, { "lr": 0.003, "gate_mid": 0.08, "gate_slope": 3.0 }, { "lr": 0.01, "gate_mid": 0.08, "gate_slope": 3.0 } ], "intervention": "analytic pendulum background, residual coordinates, defect sigmoid gate" } }