{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "mean": 3.7836247429368086e-05 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 1.23220872865204e-05 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0 }, "mean": 7.367291686932731e-06 } ], "full": { "mean": 8.597878206728637e-06, "std": 2.099193175555627e-06, "per_seed": [ 6.140434379631188e-06, 1.0502735676709563e-05, 5.883427093067439e-06, 6.942569598322734e-06, 1.1973874279647134e-05, 9.629780834075063e-06, 7.65456206863746e-06, 1.0055641723738518e-05 ], "n": 8 } }, "idea": { "mean": 1.7339235796498542e-05, "std": 4.534217009323715e-06, "per_seed": [ 1.389457884215517e-05, 2.2433650883613154e-05, 1.2446281289157923e-05, 2.6527255613473244e-05, 1.7356553144054487e-05, 1.4111134078120813e-05, 1.7410118744010106e-05, 1.453431377740344e-05 ], "n": 8 }, "comparison": { "delta_mean": 8.741357589769905e-06, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 7.754144462523982e-06, 1.1930915206903592e-05, 6.562854196090484e-06, 1.958468601515051e-05, 5.382678864407353e-06, 4.481353244045749e-06, 9.755556675372645e-06, 4.478672053664923e-06 ], "p_value": 0.0081, "mde": 4.264723393974663e-06, "mde_rel_pct": 49.60204473049084, "verdict": "idea worse (significant)", "system_worked": false, "selected_idea_cfg": { "lr": 0.006, "weight_decay": 0.0 } }, "mechanism_signature": { "phase_chaining": { "idea_by_seed": [ { "seed": 0, "high_jump_mse": 1.945050280482974e-05, "low_jump_mse": 1.2042640264553484e-05 }, { "seed": 1, "high_jump_mse": 1.8415912563796155e-05, "low_jump_mse": 2.3772785425535403e-05 }, { "seed": 2, "high_jump_mse": 1.0926943105005194e-05, "low_jump_mse": 1.2952825272805057e-05 }, { "seed": 3, "high_jump_mse": 2.6379935661680065e-05, "low_jump_mse": 2.6576328309602104e-05 }, { "seed": 4, "high_jump_mse": 1.1433454346843064e-05, "low_jump_mse": 1.933102794282604e-05 }, { "seed": 5, "high_jump_mse": 1.3358734577195719e-05, "low_jump_mse": 1.4361941794049926e-05 }, { "seed": 6, "high_jump_mse": 1.700813118077349e-05, "low_jump_mse": 1.75441091414541e-05 }, { "seed": 7, "high_jump_mse": 1.2742822946165688e-05, "low_jump_mse": 1.5131462532735895e-05 } ], "baseline_by_seed": [ { "seed": 0, "high_jump_mse": 9.345589205622673e-06, "low_jump_mse": 5.072065960121108e-06 }, { "seed": 1, "high_jump_mse": 9.975614375434816e-06, "low_jump_mse": 1.0678537364583462e-05 }, { "seed": 2, "high_jump_mse": 5.6180460887844674e-06, "low_jump_mse": 5.971849532215856e-06 }, { "seed": 3, "high_jump_mse": 7.187292339949636e-06, "low_jump_mse": 6.860935627628351e-06 }, { "seed": 4, "high_jump_mse": 1.1308992725389544e-05, "low_jump_mse": 1.2195641829748638e-05 }, { "seed": 5, "high_jump_mse": 8.861595233611297e-06, "low_jump_mse": 9.885759936878458e-06 }, { "seed": 6, "high_jump_mse": 8.615999831818044e-06, "low_jump_mse": 7.334077508858172e-06 }, { "seed": 7, "high_jump_mse": 9.407116522197612e-06, "low_jump_mse": 1.0271965038555209e-05 } ], "idea_high_jump_mse": 1.621455464828614e-05, "baseline_high_jump_mse": 8.790030790351011e-06, "idea_low_jump_mse": 1.771414008544525e-05, "baseline_low_jump_mse": 8.533854099823657e-06, "high_jump_delta_idea_minus_baseline": 7.424523857935128e-06, "idea_params": 7137, "baseline_params": 13313, "confirmed": false, "prediction": "phase chaining should reduce error on abrupt control changes" }, "idea_cfg": { "lr": 0.006, "weight_decay": 0.0 }, "idea_grid": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "result": { "mean": 9.437663175049238e-05, "std": 2.1191296338479005e-05, "per_seed": [ 0.00011513893696246669, 7.23859848221764e-05, 6.687596760457382e-05, 8.881815301720053e-05, 0.00011006097338395193, 7.692975486861542e-05, 0.00013098293857183307, 9.382034477312118e-05 ], "n": 8 } }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "result": { "mean": 3.437740951994783e-05, "std": 8.875176143263352e-06, "per_seed": [ 3.745459980564192e-05, 3.1672490877099335e-05, 2.229700839961879e-05, 3.422219378990121e-05, 3.6611334508052096e-05, 2.5733035727171227e-05, 5.4019608796807006e-05, 3.3009004255291075e-05 ], "n": 8 } }, { "cfg": { "lr": 0.006, "weight_decay": 0.0 }, "result": { "mean": 1.7339235796498542e-05, "std": 4.534217009323715e-06, "per_seed": [ 1.389457884215517e-05, 2.2433650883613154e-05, 1.2446281289157923e-05, 2.6527255613473244e-05, 1.7356553144054487e-05, 1.4111134078120813e-05, 1.7410118744010106e-05, 1.453431377740344e-05 ], "n": 8 } } ], "structural_match": "controlled pendulum rollout with an observed midpoint control event; primary metric is test MSE" } }