{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.004987949621863663 }, { "cfg": { "lr": 0.0015, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.00498378137126565 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.00358188062091358 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.0035488905268721282 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.0019017510931007564 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.0, "target_lambda": -0.08 }, "mean": 0.001846620871219784 } ], "full": { "mean": 0.0020994337101001292, "std": 0.0003699326898898649, "per_seed": [ 0.0021909112110733986, 0.0014349172124639153, 0.0016000064788386226, 0.0021606485825031996, 0.0026131158228963614, 0.002182674128562212, 0.002174227498471737, 0.0024389687459915876 ], "n": 8 } }, "idea": { "mean": 0.0019659589452203363, "std": 0.0007528780037898819, "per_seed": [ 0.002175788627937436, 0.0017074780771508813, 0.0010215449146926403, 0.001942452392540872, 0.002001833636313677, 0.0019844763446599245, 0.0012087502982467413, 0.003685347270220518 ], "n": 8 }, "comparison": { "delta_mean": -0.00013347476487979293, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ -1.5122583135962486e-05, 0.00027256086468696594, -0.0005784615641459823, -0.00021819618996232748, -0.0006112821865826845, -0.00019819778390228748, -0.0009654772002249956, 0.0012463785242289305 ], "p_value": 0.59995, "mde": 0.0005664705121806638, "mde_rel_pct": 26.982062327352402, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "prediction": "trained hidden pullback diameter contracts exponentially under shared replay", "config": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.001, "target_lambda": -0.08 }, "predicted_log_diameter_step": -0.08, "observed_log_diameter_slope": -0.4773993907385317, "final_diameter": 0.034216802567243576, "confirmed": false }, "idea_sweep": [ { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.001, "target_lambda": -0.08 }, "result": { "mean": 0.0019659589452203363, "std": 0.0007528780037898819, "per_seed": [ 0.002175788627937436, 0.0017074780771508813, 0.0010215449146926403, 0.001942452392540872, 0.002001833636313677, 0.0019844763446599245, 0.0012087502982467413, 0.003685347270220518 ], "n": 8 } }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.01, "target_lambda": -0.08 }, "result": { "mean": 0.0019659589452203363, "std": 0.0007528780037898819, "per_seed": [ 0.002175788627937436, 0.0017074780771508813, 0.0010215449146926403, 0.001942452392540872, 0.002001833636313677, 0.0019844763446599245, 0.0012087502982467413, 0.003685347270220518 ], "n": 8 } }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "epochs": 12, "attr_weight": 0.03, "target_lambda": -0.04 }, "result": { "mean": 0.0019659589452203363, "std": 0.0007528780037898819, "per_seed": [ 0.002175788627937436, 0.0017074780771508813, 0.0010215449146926403, 0.001942452392540872, 0.002001833636313677, 0.0019844763446599245, 0.0012087502982467413, 0.003685347270220518 ], "n": 8 } } ], "track_justification": "Dynamics is structurally matched: the idea diagnoses stability and contraction of recurrent latent trajectories." }