{ "bench_version": 1, "track": "tabular", "model": "local_residual_mlp", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "sweep": [ { "cfg": { "lr": 0.001, "epochs": 18, "exponent": 0.5 }, "mean": 7.946774959564209 }, { "cfg": { "lr": 0.001, "epochs": 18, "exponent": 0.8 }, "mean": 10.225590229034424 }, { "cfg": { "lr": 0.003, "epochs": 18, "exponent": 0.5 }, "mean": 7.576857328414917 }, { "cfg": { "lr": 0.003, "epochs": 18, "exponent": 0.8 }, "mean": 7.316884994506836 }, { "cfg": { "lr": 0.006, "epochs": 18, "exponent": 0.5 }, "mean": 6.160425186157227 }, { "cfg": { "lr": 0.006, "epochs": 18, "exponent": 0.8 }, "mean": 7.347582578659058 } ], "best_cfg": { "lr": 0.006, "epochs": 18, "exponent": 0.5 }, "full": { "cfg": { "lr": 0.006, "epochs": 18, "exponent": 0.5 }, "per_seed": [ 5.471584320068359, 6.244831562042236, 7.435494422912598, 5.489790439605713, 7.0974907875061035, 5.940451622009277, 7.106790065765381, 7.513341903686523 ], "mean": 6.537471890449524, "std": 0.7962674958223888, "n": 8 } }, "idea": { "cfg": { "lr": 0.006, "epochs": 18, "exponent": 0.7 }, "mean": 7.036507606506348, "per_seed": [ 7.978086948394775, 6.377083778381348, 7.57796049118042, 5.985846996307373, 7.230166435241699, 6.412862300872803, 7.17480993270874, 7.555243968963623 ], "std": 0.6550880229661826, "n": 8 }, "comparison": { "delta_mean": 0.49903571605682373, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 2.506502628326416, 0.13225221633911133, 0.14246606826782227, 0.49605655670166016, 0.1326756477355957, 0.4724106788635254, 0.06801986694335938, 0.04190206527709961 ], "p_value": 0.0081, "mde": 0.6939341374663044, "mde_rel_pct": 10.61471695931971, "verdict": "idea worse (significant)", "system_worked": false }, "math_check": { "theory_H": 0.8, "observed_H": 0.8212929503641088, "ns": [ 16, 32, 64, 128, 256 ], "scaled_sum_rms": [ 1.3914831618943797, 1.2675475655440305, 1.4158369110940938, 1.2996234839260772, 1.4794503561406334 ], "rms_ratio_last_first": 1.0632182958840077, "passed": true }, "protocol_note": "Matched tabular initialization track: same residual MLP, data, Adam, epochs, batch, and lr union.", "mechanism_signature": { "prediction": "long-memory gates retain positive adjacent correlation; critical lambda keeps scaled sums O(1)", "predicted_adjacent_corr": 0.757858283255199, "observed_adjacent_corr_mean": 0.03618130783796214, "observations": [ { "seed": 0, "baseline_adjacent_gate_corr": -0.0724730360728216, "idea_adjacent_gate_corr": 0.017231813071073975, "baseline_test_pred_std": 4.221856117248535, "idea_test_pred_std": 3.7397565841674805, "baseline_final_train_loss": 6.201660385131836, "idea_final_train_loss": 6.991382598876953 }, { "seed": 1, "baseline_adjacent_gate_corr": 0.01320452253024341, "idea_adjacent_gate_corr": 0.45370841081914715, "baseline_test_pred_std": 4.062014579772949, "idea_test_pred_std": 4.043881416320801, "baseline_final_train_loss": 4.762135734558106, "idea_final_train_loss": 5.053467597961426 }, { "seed": 2, "baseline_adjacent_gate_corr": 0.20420422523092288, "idea_adjacent_gate_corr": 0.6252111773933212, "baseline_test_pred_std": 4.786465644836426, "idea_test_pred_std": 4.8650898933410645, "baseline_final_train_loss": 3.884450912475586, "idea_final_train_loss": 4.2415578651428225 }, { "seed": 3, "baseline_adjacent_gate_corr": -0.7082938951127355, "idea_adjacent_gate_corr": -0.6668175384417161, "baseline_test_pred_std": 4.167241096496582, "idea_test_pred_std": 4.061396598815918, "baseline_final_train_loss": 6.076703529357911, "idea_final_train_loss": 6.359731369018554 }, { "seed": 4, "baseline_adjacent_gate_corr": -0.7208059938362777, "idea_adjacent_gate_corr": -0.6612016062075473, "baseline_test_pred_std": 3.6289286613464355, "idea_test_pred_std": 3.551570415496826, "baseline_final_train_loss": 5.892031574249268, "idea_final_train_loss": 6.300574836730957 }, { "seed": 5, "baseline_adjacent_gate_corr": 0.26950874374172223, "idea_adjacent_gate_corr": 0.2965993789633692, "baseline_test_pred_std": 4.365800380706787, "idea_test_pred_std": 4.453507900238037, "baseline_final_train_loss": 4.534175910949707, "idea_final_train_loss": 4.923007583618164 }, { "seed": 6, "baseline_adjacent_gate_corr": -0.3201237467772104, "idea_adjacent_gate_corr": -0.32051733449908076, "baseline_test_pred_std": 3.79843807220459, "idea_test_pred_std": 3.9083645343780518, "baseline_final_train_loss": 7.1678547668457036, "idea_final_train_loss": 7.365681228637695 }, { "seed": 7, "baseline_adjacent_gate_corr": 0.4083689217021288, "idea_adjacent_gate_corr": 0.5452361616051298, "baseline_test_pred_std": 4.7109575271606445, "idea_test_pred_std": 4.555789947509766, "baseline_final_train_loss": 3.857210865020752, "idea_final_train_loss": 4.864171485900879 } ], "confirmed": false } }