{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.006, "weight_decay": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0 }, "mean": 0.005481100233737379 }, { "cfg": { "lr": 0.0015, "weight_decay": 0.0001 }, "mean": 0.005474042263813317 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.003975238680141047 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.00394076865632087 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0 }, "mean": 0.002146739774616435 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001 }, "mean": 0.002088654466206208 } ], "full": { "mean": 0.0022117759071988985, "std": 0.0004126881098400608, "per_seed": [ 0.002800536109134555, 0.001347046927548945, 0.001825663261115551, 0.0023813715670257807, 0.0023169058840721846, 0.0024658460170030594, 0.0022247303277254105, 0.002332107163965702 ], "n": 8 } }, "idea": { "mean": 0.002208090227213688, "std": 0.00051193081082026, "per_seed": [ 0.002686932682991028, 0.001070088124834001, 0.002036568708717823, 0.002309913281351328, 0.0025295668747276068, 0.0025805803015828133, 0.0026061630342155695, 0.0018449088092893362 ], "n": 8, "best_cfg": { "lr": 0.006, "weight_decay": 0.0001, "alpha": 0.03 }, "trials": [ { "cfg": { "lr": 0.0015, "weight_decay": 0.0001, "alpha": 0.03 }, "mean": 0.005427066018455662 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "alpha": 0.03 }, "mean": 0.003149385840515606 }, { "cfg": { "lr": 0.006, "weight_decay": 0.0001, "alpha": 0.03 }, "mean": 0.002208090227213688 } ] }, "comparison": { "delta_mean": -3.6856799852102995e-06, "idea_wins": 4, "n_pairs": 8, "per_seed_diffs": [ -0.00011360342614352703, -0.0002769588027149439, 0.00021090544760227203, -7.145828567445278e-05, 0.0002126609906554222, 0.00011473428457975388, 0.00038143270649015903, -0.00048719835467636585 ], "p_value": 0.97695, "mde": 0.00024075513247358545, "mde_rel_pct": 10.88515033055449, "verdict": "no measurable effect", "system_worked": false }, "paired_deltas": [ -0.00011360342614352703, -0.0002769588027149439, 0.00021090544760227203, -7.145828567445278e-05, 0.0002126609906554222, 0.00011473428457975388, 0.00038143270649015903, -0.00048719835467636585 ], "permutation_p": 0.97695, "mechanism_signature": { "q_observed": 0.010351310360661614, "q_predicted": 0.010351310360661614, "drift_energy_ratio": 1.0, "orthogonality_residual": 0.0, "prediction": "cumulative quadratic dissipation is capped at QMAX and skew drift is orthogonal to the instantaneous gradient", "qmax": 0.5, "confirmed": true } }