{ "bench_version": 1, "track": "tabular", "model": "mlp_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "sweep": { "best_cfg": { "lr": 0.009 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 15.798457145690918 }, { "cfg": { "lr": 0.003 }, "mean": 9.297731161117554 }, { "cfg": { "lr": 0.009 }, "mean": 6.843065023422241 } ], "full": { "mean": 6.911551356315613, "std": 0.25102731009725415, "per_seed": [ 6.589434623718262, 6.794177055358887, 7.121844291687012, 6.866804122924805, 6.845246315002441, 7.470037460327148, 6.813384056091309, 6.791482925415039 ], "n": 8 } }, "best_config": { "lr": 0.009, "epochs": 15 }, "full": { "mean": 6.911551356315613, "std": 0.25102731009725415, "per_seed": [ 6.589434623718262, 6.794177055358887, 7.121844291687012, 6.866804122924805, 6.845246315002441, 7.470037460327148, 6.813384056091309, 6.791482925415039 ], "n": 8 } }, "idea": { "config": { "lr": 0.009, "epochs": 15, "R": 2.5, "block": "first_layer_bias", "metric": "mse" }, "per_seed": [ 6.712672233581543, 6.749656677246094, 6.953096389770508, 6.8154120445251465, 6.669680118560791, 7.3650336265563965, 6.754833221435547, 6.995800971984863 ], "tuning_sweep": [ { "lr": 0.001, "mean_mse": 15.960896253585815, "per_seed": [ 15.142996788024902, 15.989051818847656, 16.95168685913086, 15.759849548339844 ] }, { "lr": 0.003, "mean_mse": 8.806230664253235, "per_seed": [ 7.58334493637085, 9.324054718017578, 9.102656364440918, 9.214866638183594 ] }, { "lr": 0.009, "mean_mse": 6.807709336280823, "per_seed": [ 6.712672233581543, 6.749656677246094, 6.953096389770508, 6.8154120445251465 ] } ], "mean": 6.877023160457611, "std": 0.2130142060373052, "n": 8 }, "comparison": { "delta_mean": -0.03452819585800171, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ 0.12323760986328125, -0.04452037811279297, -0.1687479019165039, -0.0513920783996582, -0.1755661964416504, -0.10500383377075195, -0.05855083465576172, 0.20431804656982422 ], "p_value": 0.46775, "mde": 0.11208041708387258, "mde_rel_pct": 1.6216390692294593, "verdict": "no measurable effect", "system_worked": false }, "math_check": { "max_intrinsic_grad_g_squared": 0.9999989999995, "max_bound_violation": 0.0, "max_radial_formula_error": 3.3306690738754696e-16, "step_inverse_metric_ratio_r0.9999": 0.0006689383838711185 }, "selection_note": "baseline and idea share lr union; baseline tuned on seeds 0-3; full comparison uses paired seeds 0-7", "runtime_sec": 30.761850833892822, "mechanism_signature": { "prediction": "inverse barrier metric decreases toward boundary; bounded intrinsic barrier gradient", "predicted_at_center": 3.125, "observed_mean_inverse_metric_at_trained_points": 1.705785185098648, "trained_model_points": [ { "seed": 0, "observed_mean_inverse_metric": 1.6924855709075928, "final_q": 0.5548654556274414 }, { "seed": 1, "observed_mean_inverse_metric": 1.7231124639511108, "final_q": 0.5643916320800781 }, { "seed": 2, "observed_mean_inverse_metric": 1.6797361373901367, "final_q": 0.550905647277832 }, { "seed": 3, "observed_mean_inverse_metric": 1.7380695343017578, "final_q": 0.5688753509521485 }, { "seed": 4, "observed_mean_inverse_metric": 1.864469289779663, "final_q": 0.6083690643310546 }, { "seed": 5, "observed_mean_inverse_metric": 1.7799270153045654, "final_q": 0.5821066665649415 }, { "seed": 6, "observed_mean_inverse_metric": 1.812181830406189, "final_q": 0.592091178894043 }, { "seed": 7, "observed_mean_inverse_metric": 1.356299638748169, "final_q": 0.44984680175781255 } ], "confirmed": true } }