{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "wd": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.001, "wd": 0.0 }, "mean": 0.004538182052783668 }, { "cfg": { "lr": 0.001, "wd": 0.0001 }, "mean": 0.004533302329946309 }, { "cfg": { "lr": 0.003, "wd": 0.0 }, "mean": 0.0035770683316513896 }, { "cfg": { "lr": 0.003, "wd": 0.0001 }, "mean": 0.0035493404720909894 }, { "cfg": { "lr": 0.006, "wd": 0.0 }, "mean": 0.0014125758316367865 }, { "cfg": { "lr": 0.006, "wd": 0.0001 }, "mean": 0.0013684789882972836 } ], "full": { "mean": 0.0015742218820378184, "std": 0.0005383299843650786, "per_seed": [ 0.0017803608207032084, 0.001374608720652759, 0.0006994276773184538, 0.0016195187345147133, 0.0015369732864201069, 0.0012765798019245267, 0.0027483156882226467, 0.0015579903265461326 ], "n": 8 } }, "idea": { "mean": 0.0015715822883066721, "std": 0.0005392184858503133, "per_seed": [ 0.001779490732587874, 0.0013879159232601523, 0.0007163107511587441, 0.0014804680831730366, 0.0015849629417061806, 0.0012773972703143954, 0.002764256903901696, 0.0015818557003512979 ], "n": 8 }, "comparison": { "delta_mean": -2.639593731146306e-06, "idea_wins": 2, "n_pairs": 8, "per_seed_diffs": [ -8.700881153345108e-07, 1.3307202607393265e-05, 1.6883073840290308e-05, -0.0001390506513416767, 4.7989655286073685e-05, 8.174683898687363e-07, 1.5941215679049492e-05, 2.386537380516529e-05 ], "p_value": 0.985, "mde": 4.7793207118518294e-05, "mde_rel_pct": 3.0359892505528094, "verdict": "no measurable effect", "system_worked": false }, "track_justification": "dynamics is the matched actuated-pendulum rollout track for stability/control ideas", "idea_sweep": [ { "cfg": { "lr": 0.006, "wd": 0.0001, "phase": 0.0, "tv": 0.0 }, "mean": 0.0013684789882972836 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "phase": 0.002, "tv": 0.01 }, "mean": 0.0013676858798135072 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "phase": 0.008, "tv": 0.04 }, "mean": 0.0013410463725449517 } ], "math_check": { "expected_sum_theta_dot": -0.15, "finite_difference_q": -0.15000000002777925, "absolute_error": 2.7779251121629045e-11, "passed": true }, "custom_track": null, "mechanism_signature": { "quantity": "test hidden angular-velocity variance q", "prediction": "phase budget lowers q variance", "baseline_mean": 0.06730567710474133, "idea_mean": 0.06802853476256132, "ratio_idea_over_baseline": 1.0107399210425456, "relative_reduction": -0.01073992104254562, "confirmed": false } }