{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 0.0014904916170053184 }, { "cfg": { "lr": 0.003 }, "mean": 0.0010051239951280877 }, { "cfg": { "lr": 0.01 }, "mean": 0.00037887565849814564 } ], "full": { "mean": 0.0003780504302994814, "std": 7.314250470278514e-05, "per_seed": [ 0.0003051091334782541, 0.00044008868280798197, 0.0002926669840235263, 0.0004776378336828202, 0.00027556720306165516, 0.00042108612251468003, 0.00037083675852045417, 0.0004414107243064791 ], "n": 8 } }, "idea": { "mean": 0.00030132598112686537, "std": 6.586114249212311e-05, "per_seed": [ 0.00032067738357000053, 0.00030712690204381943, 0.00022303542937152088, 0.0003347629099152982, 0.000343360414262861, 0.00025801677838899195, 0.00020414931350387633, 0.0004194787179585546 ], "n": 8 }, "comparison": { "delta_mean": -7.6724449172616e-05, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ 1.556825009174645e-05, -0.00013296178076416254, -6.963155465200543e-05, -0.00014287492376752198, 6.779321120120585e-05, -0.00016306934412568808, -0.00016668744501657784, -2.193200634792447e-05 ], "p_value": 0.0522, "mde": 7.448216191901723e-05, "mde_rel_pct": 19.701647174429734, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "track_match": "dynamics/control", "intervention": "IQC-inspired certified gradient clipping", "idea_sweep": [ { "cfg": { "lr": 0.001, "clip": 1.0 }, "mean": 0.001604030971066095 }, { "cfg": { "lr": 0.003, "clip": 1.0 }, "mean": 0.0008869413213687949 }, { "cfg": { "lr": 0.01, "clip": 1.0 }, "mean": 0.00030132598112686537 } ], "best_idea_cfg": { "lr": 0.01, "clip": 1.0 }, "signature": { "predicted_gamma_proxy": 12.570822464322077, "observed_output_gain": 7.226121114322568e-05, "disturbance_epsilon": 5.01467227935791, "output_shift": 0.000362366292392835, "parameter_shift": 0.03903782367706299, "estimated_incremental_gain_L": 0.08642427995792329, "rho": 0.9204507101394397, "clip_events": 6, "confirmed": true } }, "protocol_notes": { "epochs": 12, "batch": 64, "n_train": 400, "n_test": 200, "baseline_grid": [ { "lr": 0.001 }, { "lr": 0.003 }, { "lr": 0.01 } ], "idea_grid": [ { "lr": 0.001, "clip": 1.0 }, { "lr": 0.003, "clip": 1.0 }, { "lr": 0.01, "clip": 1.0 } ], "paired_seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ] } }