{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0, "epochs": 8 }, "mean": 0.02264283667318523 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001, "epochs": 8 }, "mean": 0.022735907114110887 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0, "epochs": 8 }, "mean": 0.006244247197173536 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "epochs": 8 }, "mean": 0.00622874335385859 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0, "epochs": 8 }, "mean": 0.00514151350944303 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8 }, "mean": 0.005085793818579987 } ], "full": { "mean": 0.004114087234484032, "std": 0.0026467894721201165, "per_seed": [ 0.0021156652364879847, 0.008485781028866768, 0.0016995229525491595, 0.008042206056416035, 0.0027983859181404114, 0.001188579830341041, 0.003477072110399604, 0.005105484742671251 ], "n": 8 } }, "idea": { "mean": 0.004023682136903517, "std": 0.0026034026575959125, "per_seed": [ 0.002100873738527298, 0.008452225476503372, 0.0018759106751531363, 0.007780176121741533, 0.002418885938823223, 0.0011175131658092141, 0.0034842072054743767, 0.004959664773195982 ], "n": 8 }, "comparison": { "delta_mean": -9.040509758051485e-05, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ -1.4791497960686684e-05, -3.355555236339569e-05, 0.00017638772260397673, -0.0002620299346745014, -0.00037949997931718826, -7.106666453182697e-05, 7.135095074772835e-06, -0.00014581996947526932 ], "p_value": 0.1825, "mde": 0.00014348105739960172, "mde_rel_pct": 3.4875550570963134, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "idea_config": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8, "rho": 0.05 }, "idea_sweep": [ { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8, "rho": 0.001 }, "mean": 0.004106936205062084 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8, "rho": 0.01 }, "mean": 0.004095953670912422 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001, "epochs": 8, "rho": 0.05 }, "mean": 0.004023682136903517 } ], "prediction": "BT monitor drives delta,tau toward zero and the trained local dynamics toward a near-zero continuous eigenvalue", "trained_mean_abs_delta": 0.08188444655388594, "trained_mean_abs_tau": 0.573352038860321, "observed_mean_discrete_radius": 0.736884132027626, "observed_mean_continuous_max_real": -0.26311645470559597, "confirmed": false }, "custom_track": null }