{ "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, "epochs": 3 }, "sweep": [ { "cfg": { "lr": 0.0015, "wd": 0.0, "epochs": 3 }, "mean": 0.3753514513373375 }, { "cfg": { "lr": 0.0015, "wd": 0.0001, "epochs": 3 }, "mean": 0.37542054802179337 }, { "cfg": { "lr": 0.003, "wd": 0.0, "epochs": 3 }, "mean": 0.02805777284083888 }, { "cfg": { "lr": 0.003, "wd": 0.0001, "epochs": 3 }, "mean": 0.02808422112138942 }, { "cfg": { "lr": 0.006, "wd": 0.0, "epochs": 3 }, "mean": 0.019103779457509518 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 3 }, "mean": 0.019099895376712084 } ], "full": { "mean": 0.021109665918629616, "std": 0.013601078749505634, "per_seed": [ 0.03851161524653435, 0.010536076501011848, 0.011643605306744576, 0.015708284452557564, 0.045241355895996094, 0.007054212968796492, 0.011020926758646965, 0.029161250218749046 ], "n": 8 } }, "idea": { "mean": 0.02110744162928313, "std": 0.013599407708771062, "per_seed": [ 0.03850656375288963, 0.010544697754085064, 0.011639888398349285, 0.01570560224354267, 0.045238953083753586, 0.007051908411085606, 0.011017857119441032, 0.029154062271118164 ], "n": 8, "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 3, "lam": 0.02, "tau": 1.0 } }, "comparison": { "delta_mean": -2.22428934648633e-06, "idea_wins": 7, "n_pairs": 8, "per_seed_diffs": [ -5.0514936447143555e-06, 8.621253073215485e-06, -3.7169083952903748e-06, -2.682209014892578e-06, -2.4028122425079346e-06, -2.3045577108860016e-06, -3.069639205932617e-06, -7.187947630882263e-06 ], "p_value": 0.21825, "mde": 3.914864826646565e-06, "mde_rel_pct": 0.018545366097867208, "verdict": "no measurable effect", "system_worked": false }, "mechanism_signature": { "prediction": "Training the positive-cycle penalty suppresses measured short closed-walk gain in the recurrent hidden Jacobian.", "observed_baseline_mean_trace2": 0.002615979639813304, "observed_idea_mean_trace2": 0.0026157096144743264, "relative_change_pct": -0.010322149869512742, "math_sanity": { "trace2": 2.8800000000000003, "expected_2ab": 2.8800000000000003, "abs_error": 0.0, "rho": 1.2000000000000002, "critical_gain": 0.8333333333333333 }, "confirmed": true }, "idea_sweep": [ { "cfg": { "lr": 0.0015, "wd": 0.0, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.33826782926917076 }, { "cfg": { "lr": 0.0015, "wd": 0.0, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.3382681477814913 }, { "cfg": { "lr": 0.0015, "wd": 0.0, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.3382690101861954 }, { "cfg": { "lr": 0.0015, "wd": 0.0001, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.33832092210650444 }, { "cfg": { "lr": 0.0015, "wd": 0.0001, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.3383213337510824 }, { "cfg": { "lr": 0.0015, "wd": 0.0001, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.33832169882953167 }, { "cfg": { "lr": 0.003, "wd": 0.0, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.032877288351301104 }, { "cfg": { "lr": 0.003, "wd": 0.0, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.032877395395189524 }, { "cfg": { "lr": 0.003, "wd": 0.0, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.03287789586465806 }, { "cfg": { "lr": 0.003, "wd": 0.0001, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.03286271571414545 }, { "cfg": { "lr": 0.003, "wd": 0.0001, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.03286272965488024 }, { "cfg": { "lr": 0.003, "wd": 0.0001, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.03286092990310863 }, { "cfg": { "lr": 0.006, "wd": 0.0, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.02112848818069324 }, { "cfg": { "lr": 0.006, "wd": 0.0, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.021127756568603218 }, { "cfg": { "lr": 0.006, "wd": 0.0, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.021124891936779022 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 3, "lam": 0.001, "tau": 1.0 }, "mean": 0.021109436987899244 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 3, "lam": 0.005, "tau": 1.0 }, "mean": 0.02110862295376137 }, { "cfg": { "lr": 0.006, "wd": 0.0001, "epochs": 3, "lam": 0.02, "tau": 1.0 }, "mean": 0.02110744162928313 } ] }