{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 0.0012379240797599778 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001 }, "mean": 0.001226543157827109 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.000877063677762635 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.0008647404611110687 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "mean": 0.0003551548215909861 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001 }, "mean": 0.00048234833229798824 } ], "full": { "mean": 0.0002390247773291776, "std": 0.00016356887817964183, "per_seed": [ 0.000364280364010483, 0.0005520923878066242, 0.0003466835187282413, 0.00015756301581859589, 0.00026062532560899854, 0.00010608434968162328, 3.411849320400506e-05, 9.075076377484947e-05 ], "n": 8 } }, "idea": { "mean": 0.7325335219502449, "std": 0.07627789337540042, "per_seed": [ 0.6195922493934631, 0.8263598680496216, 0.7263849377632141, 0.8282962441444397, 0.634881854057312, 0.7560707926750183, 0.681036651134491, 0.7876455783843994 ], "n": 8 }, "comparison": { "delta_mean": 0.7322944971729157, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 0.6192279690294527, 0.825807775661815, 0.7260382542444859, 0.8281386811286211, 0.634621228731703, 0.7559647083253367, 0.681002532641287, 0.7875548276206246 ], "p_value": 0.0081, "mde": 0.06817817350226507, "mde_rel_pct": 28523.47537526295, "verdict": "idea worse (significant)", "system_worked": false }, "mechanism_signature": { "prediction": "parameter updates become active after q>gamma and loss then decreases", "predicted_vs_observed": { "predicted_positive_q_activation": true, "observed_positive_q_rate_mean": 0.0, "observed_activation_rate_mean": 0.0, "observed_post_activation_loss_drop_mean": 0.0, "activation_epoch_values": [ null, null, null, null, null, null, null, null ] }, "confirmed": false }, "idea_sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248376846 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248376846 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248376846 } ], "protocol": { "seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "n_train": 900, "n_test": 300, "epochs": 14, "structural_match": "controlled pendulum rollout / latent dynamics", "custom_track": null } }