{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.001, "epochs": 12 }, "sweep": [ { "cfg": { "lr": 0.001, "epochs": 12 }, "per_seed": [ 0.029528150334954262, 0.04754278063774109, 0.033632513135671616, 0.031397681683301926, 0.029850929975509644, 0.12557671964168549, 0.022646039724349976, 0.05035697668790817 ], "mean": 0.04631647397764027, "std": 0.031218425382336913, "n": 8 }, { "cfg": { "lr": 0.003, "epochs": 12 }, "per_seed": [ 0.05427622050046921, 0.028770923614501953, 0.41897544264793396, 0.025073563680052757, 0.07675475627183914, 0.16374778747558594, 0.06638805568218231, 0.09212644398212433 ], "mean": 0.1157641492318362, "std": 0.12161998162703432, "n": 8 }, { "cfg": { "lr": 0.01, "epochs": 12 }, "per_seed": [ 0.68760746717453, 0.04096516966819763, 0.026457389816641808, 0.32609719038009644, 0.01830579899251461, 0.08879950642585754, 0.2804693281650543, 0.6182840466499329 ], "mean": 0.26087323715910316, "std": 0.25155552050299584, "n": 8 } ], "harness_tuning": { "best_cfg": { "lr": 0.001, "epochs": 12 }, "sweep": [ { "cfg": { "lr": 0.001, "epochs": 12 }, "mean": 0.03552528144791722 }, { "cfg": { "lr": 0.003, "epochs": 12 }, "mean": 0.13177403761073947 }, { "cfg": { "lr": 0.01, "epochs": 12 }, "mean": 0.2702818042598665 } ], "full": { "mean": 0.04631647397764027, "std": 0.031218425382336913, "per_seed": [ 0.029528150334954262, 0.04754278063774109, 0.033632513135671616, 0.031397681683301926, 0.029850929975509644, 0.12557671964168549, 0.022646039724349976, 0.05035697668790817 ], "n": 8 } }, "full": { "per_seed": [ 0.029528150334954262, 0.04754278063774109, 0.033632513135671616, 0.031397681683301926, 0.029850929975509644, 0.12557671964168549, 0.022646039724349976, 0.05035697668790817 ], "mean": 0.04631647397764027, "std": 0.031218425382336913, "n": 8 } }, "idea": { "per_seed": [ 0.029486622661352158, 0.047586631029844284, 0.03362056985497475, 0.03137165307998657, 0.02987625077366829, 0.11335577070713043, 0.022612692788243294, 0.05035630613565445 ], "mean": 0.04478331212885678, "std": 0.0273708875835822, "n": 8 }, "comparison": { "delta_mean": -0.001533161848783493, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ -4.152767360210419e-05, 4.385039210319519e-05, -1.1943280696868896e-05, -2.6028603315353394e-05, 2.532079815864563e-05, -0.012220948934555054, -3.3346936106681824e-05, -6.705522537231445e-07 ], "p_value": 0.2886, "mde": 0.003611027360643652, "mde_rel_pct": 7.796421123047731, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "idea_sweep": [ { "cfg": { "lr": 0.001, "epochs": 12 }, "per_seed": [ 0.029486622661352158, 0.047586631029844284, 0.03362056985497475, 0.03137165307998657, 0.02987625077366829, 0.11335577070713043, 0.022612692788243294, 0.05035630613565445 ], "mean": 0.04478331212885678, "std": 0.0273708875835822, "n": 8 }, { "cfg": { "lr": 0.003, "epochs": 12 }, "per_seed": [ 0.054160747677087784, 0.14119376242160797, 0.014620370231568813, 0.03189314156770706, 0.04598933085799217, 0.3756285011768341, 0.04325627163052559, 0.07682377099990845 ], "mean": 0.097945737070404, "std": 0.110870329571024, "n": 8 }, { "cfg": { "lr": 0.01, "epochs": 12 }, "per_seed": [ 0.9418538212776184, 0.0459330677986145, 0.025276441127061844, 0.38526275753974915, 0.034148477017879486, 0.09575814008712769, 0.09502670913934708, 6.104557991027832 ], "mean": 0.9659771756269038, "std": 1.963999860256041, "n": 8 } ], "paired_deltas_idea_minus_baseline": [ -4.152767360210419e-05, 4.385039210319519e-05, -1.1943280696868896e-05, -2.6028603315353394e-05, 2.532079815864563e-05, -0.012220948934555054, -3.3346936106681824e-05, -6.705522537231445e-07 ], "permutation_pvalue": 0.2886, "signature": { "prediction": "phase certificate excess is suppressed by the Jacobian penalty", "predicted_excess_values": [ 0.8990192516358979, 0.941527019313205, 0.8427875144990571, 0.8680509431871064 ], "observed_block_gain_proxy": [ 1.048457384109497, 1.0909651517868042, 0.9922256469726562, 1.0174890756607056 ], "trained_test_mse": 0.029486622661352158, "confirmed": true }, "track_selection": "dynamics: actuated pendulum stability/control has recurrent propagation structure." }, "custom_track": null }