{ "bench_version": 1, "track": "tabular", "model": "mlp_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 146.7281608581543 }, { "cfg": { "lr": 0.003 }, "mean": 17.104151725769043 }, { "cfg": { "lr": 0.01 }, "mean": 12.445883750915527 } ], "full": { "mean": 12.770212650299072, "std": 1.168904265972683, "per_seed": [ 12.32522201538086, 12.587058067321777, 14.008906364440918, 10.862348556518555, 11.249069213867188, 13.969073295593262, 13.09931755065918, 14.06070613861084 ], "n": 8 } }, "idea": { "mean": 14.78523874282837, "per_seed": [ 17.6944580078125, 12.148554801940918, 14.37415885925293, 26.43087387084961, 8.728538513183594, 12.932696342468262, 13.736586570739746, 12.236042976379395 ], "cfg": { "lr": 0.001 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 14.78523874282837, "per_seed": [ 17.6944580078125, 12.148554801940918, 14.37415885925293, 26.43087387084961, 8.728538513183594, 12.932696342468262, 13.736586570739746, 12.236042976379395 ] }, { "cfg": { "lr": 0.003 }, "mean": 25.065056443214417, "per_seed": [ 18.1739444732666, 15.587058067321777, 16.13030242919922, 13.475675582885742, 65.8013687133789, 8.955453872680664, 17.400615692138672, 44.99603271484375 ] }, { "cfg": { "lr": 0.01 }, "mean": 16.26691508293152, "per_seed": [ 42.245765686035156, 13.930212020874023, 9.309124946594238, 8.357898712158203, 22.75505828857422, 8.172164916992188, 16.167266845703125, 9.197829246520996 ] } ] }, "comparison": { "delta_mean": 2.015026092529297, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ 5.369235992431641, -0.4385032653808594, 0.3652524948120117, 15.568525314331055, -2.5205307006835938, -1.036376953125, 0.6372690200805664, -1.8246631622314453 ], "p_value": 0.4921, "mde": 4.999106670375855, "mde_rel_pct": 39.14662039914252, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "track_choice": "tabular/Friedman#1: optimizer intervention is structurally matched", "prediction": "small-gap positive-log lambda/delta approximates inverse weighted gradient square", "observed_lambda_over_delta_mean": 0.2710435168226066, "predicted_inverse_variance_mean": 1000000000000.0, "root_residual_max": 0.169219970703125, "root_failure_rate_mean": 0.14531249999999998, "confirmed": false } }