{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.0 }, "mean": 0.01910123834386468 }, { "cfg": { "lr": 0.003, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.0 }, "mean": 0.005347788450308144 }, { "cfg": { "lr": 0.01, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.0 }, "mean": 0.0039304872043430805 } ], "full": { "mean": 0.003924444579752162, "std": 0.0009507399174062226, "per_seed": [ 0.005481751170009375, 0.003171606920659542, 0.0028356127440929413, 0.004232977982610464, 0.004473012872040272, 0.0031893684063106775, 0.005031306762248278, 0.0029799197800457478 ], "n": 8 } }, "idea": { "mean": 0.0036855603102594614, "std": 0.0008949623394657411, "per_seed": [ 0.0049509322270751, 0.0027915509417653084, 0.0026541445404291153, 0.0039238822646439075, 0.0043672905303537846, 0.0031351130455732346, 0.004849778488278389, 0.002811790443956852 ], "n": 8 }, "comparison": { "delta_mean": -0.0002388842694927007, "idea_wins": 8, "n_pairs": 8, "per_seed_diffs": [ -0.0005308189429342747, -0.0003800559788942337, -0.000181468203663826, -0.00030909571796655655, -0.0001057223416864872, -5.425536073744297e-05, -0.0001815282739698887, -0.0001681293360888958 ], "p_value": 0.0081, "mde": 0.00013163873291981864, "mde_rel_pct": 3.3543277333816226, "verdict": "idea better (significant)", "system_worked": true }, "idea_sweep": [ { "cfg": { "lr": 0.01, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.0 }, "mean": 0.003924444579752162 }, { "cfg": { "lr": 0.01, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.03 }, "mean": 0.0036855603102594614 }, { "cfg": { "lr": 0.01, "wd": 0.0001, "epochs": 12, "batch": 128, "rank": 8, "coupling": 0.1 }, "mean": 0.003690520068630576 } ], "device": "cpu", "mechanism_signature": { "predicted": "shared disagreement decreases while private variance remains nonzero", "observed_idea_shared_disagreement_mean": 0.020418916479684412, "observed_baseline_shared_disagreement_mean": 0.06511001940816641, "observed_idea_private_variance_mean": 0.08402373269200325, "observed_baseline_private_variance_mean": 0.08042469155043364, "confirmed": true }, "custom_track": null }