{ "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "clip_margin": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "clip_margin": 0.0 }, "mean": 0.008678926387801766 }, { "cfg": { "lr": 0.001, "clip_margin": 0.03 }, "mean": 0.010034722508862615 }, { "cfg": { "lr": 0.001, "clip_margin": 0.08 }, "mean": 0.012918951804749668 }, { "cfg": { "lr": 0.003, "clip_margin": 0.0 }, "mean": 0.006641886895522475 }, { "cfg": { "lr": 0.003, "clip_margin": 0.03 }, "mean": 0.007589806220494211 }, { "cfg": { "lr": 0.003, "clip_margin": 0.08 }, "mean": 0.009543078020215034 }, { "cfg": { "lr": 0.01, "clip_margin": 0.0 }, "mean": 0.005315697111655027 }, { "cfg": { "lr": 0.01, "clip_margin": 0.03 }, "mean": 0.006531292689032853 }, { "cfg": { "lr": 0.01, "clip_margin": 0.08 }, "mean": 0.008570412872359157 } ], "full": { "mean": 0.0046574123844038695, "std": 0.0013966285668353233, "per_seed": [ 0.003739116946235299, 0.004545919597148895, 0.005411764141172171, 0.007565987762063742, 0.004889188800007105, 0.002318228594958782, 0.00456235371530056, 0.004226739518344402 ], "n": 8 } }, "idea": { "mean": 0.0047074658796191216, "std": 0.0014044730096107532, "per_seed": [ 0.003767308546230197, 0.004722419194877148, 0.005457298830151558, 0.007620518561452627, 0.004916135687381029, 0.002345199463889003, 0.004579263739287853, 0.0042515830136835575 ], "n": 8 }, "comparison": { "delta_mean": 5.005349521525204e-05, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 2.8191599994897842e-05, 0.0001764995977282524, 4.553468897938728e-05, 5.45307993888855e-05, 2.6946887373924255e-05, 2.6970868930220604e-05, 1.6910023987293243e-05, 2.4843495339155197e-05 ], "p_value": 0.0081, "mde": 4.3914032994549896e-05, "mde_rel_pct": 0.9428847903098175, "verdict": "idea worse (significant)", "system_worked": false }, "mechanism_signature": { "selected_idea_cfg": { "lr": 0.01, "kappa": 0.5 }, "baseline_best_cfg": { "lr": 0.01, "clip_margin": 0.0 }, "idea_sweep": [ { "cfg": { "lr": 0.001, "kappa": 0.5 }, "mean_4seed": 0.008722120779566467 }, { "cfg": { "lr": 0.001, "kappa": 1.0 }, "mean_4seed": 0.008760003955103457 }, { "cfg": { "lr": 0.001, "kappa": 2.0 }, "mean_4seed": 0.008845672477036715 }, { "cfg": { "lr": 0.003, "kappa": 0.5 }, "mean_4seed": 0.006680505815893412 }, { "cfg": { "lr": 0.003, "kappa": 1.0 }, "mean_4seed": 0.006711311521939933 }, { "cfg": { "lr": 0.003, "kappa": 2.0 }, "mean_4seed": 0.0067982005421072245 }, { "cfg": { "lr": 0.01, "kappa": 0.5 }, "mean_4seed": 0.005391886283177882 }, { "cfg": { "lr": 0.01, "kappa": 1.0 }, "mean_4seed": 0.005438438267447054 }, { "cfg": { "lr": 0.01, "kappa": 2.0 }, "mean_4seed": 0.005516079894732684 } ], "mechanism_signature": { "prediction": "adaptive ebar tracks observed model residual and larger kappa increases intervention", "predicted_vs_observed_mae": 1.7471262253820896e-05, "relative_mae": 0.007133921903418612, "rows": [ { "seed": 0, "predicted_ebar": 0.002444567857310176, "observed_residual": 0.0024249006528407335, "intervention": 0.007492117118090391, "test_mse": 0.003767308546230197 }, { "seed": 1, "predicted_ebar": 0.002442610915750265, "observed_residual": 0.00243778177537024, "intervention": 0.0064279804937541485, "test_mse": 0.004722419194877148 }, { "seed": 2, "predicted_ebar": 0.0024595605209469795, "observed_residual": 0.0024749869480729103, "intervention": 0.014962565153837204, "test_mse": 0.005457298830151558 }, { "seed": 3, "predicted_ebar": 0.0024496361147612333, "observed_residual": 0.002463151467964053, "intervention": 0.015951024368405342, "test_mse": 0.007620518561452627 }, { "seed": 4, "predicted_ebar": 0.002494362648576498, "observed_residual": 0.0024751098826527596, "intervention": 0.011707877740263939, "test_mse": 0.004916135687381029 }, { "seed": 5, "predicted_ebar": 0.0024386330042034388, "observed_residual": 0.002423582598567009, "intervention": 0.004593744874000549, "test_mse": 0.002345199463889003 }, { "seed": 6, "predicted_ebar": 0.002432666253298521, "observed_residual": 0.002475534798577428, "intervention": 0.005940093658864498, "test_mse": 0.004579263739287853 }, { "seed": 7, "predicted_ebar": 0.002408114029094577, "observed_residual": 0.002417274285107851, "intervention": 0.008047887124121189, "test_mse": 0.0042515830136835575 } ], "confirmed": true } } }