Diversity-Weighted Leave-One-Out Policy Baseline / report_bench_2026-09-01T132743.md
Mechanism confirmed, baseline not beaten
Стенд-проверка (stage-2) · промт оператора:
(универсальный)
Ответ агента:
{ "worked": false, "confidence": 9, "verdict": "The idea was implemented as a grouped Gaussian rollout objective on the registered dynamics track using the shared rnn_small architecture, with uniform leave-one-out as baseline and diversity-weighted leave-one-out as the only intervention. The mechanism signature was confirmed on trained-model representations, but the idea was slightly worse on test MSE and not significant, so it does not qualify as a win.", "metrics": { "baseline": "Uniform LOO, tuned lr=0.01: test MSE 0.0144065906 ± 0.0059627913 over 8 seeds.", "idea": "Diversity-weighted LOO p=1: test MSE 0.0149132591 ± 0.0066294191 over 8 seeds; delta +0.0005066685, permutation p=0.5343." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_bench.py", "files": [ "stage2_bench.py", "bench_report.json", "diversity_baseline.py", "verification.json" ], "limitations": "Only the registered dynamics track was tested because the idea requires grouped stochastic trajectories. The implementation is a Gaussian dynamics-prediction surrogate rather than a full combinatorial solver; no throughput or gradient-variance comparison was included, and the idea-side sweep used p=1 only with a duplicate nearby configuration.", "bench_report": { "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "epochs": 6, "p": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "epochs": 6, "p": 0.0 }, "mean": 0.1710126642137766 }, { "cfg": { "lr": 0.003, "epochs": 6, "p": 0.0 }, "mean": 0.024177383398637176 }, { "cfg": { "lr": 0.01, "epochs": 6, "p": 0.0 }, "mean": 0.012901934329420328 } ], "full": { "mean": 0.014406590606085956, "std": 0.005962791282569369, "per_seed": [ 0.010699832811951637, 0.012548977509140968, 0.009432518854737282, 0.018926408141851425, 0.02494051493704319, 0.01899670995771885, 0.004924832843244076, 0.014782929793000221 ], "n": 8 } }, "idea": { "mean": 0.014913259132299572, "std": 0.006629419112933174, "per_seed": [ 0.008315389044582844, 0.016753537580370903, 0.00829843245446682, 0.019920941442251205, 0.02598617970943451, 0.019167941063642502, 0.005189504008740187, 0.015674147754907608 ], "n": 8 }, "comparison": { "delta_mean": 0.0005066685262136161, "idea_wins": 2, "n_pairs": 8, "per_seed_diffs": [ -0.0023844437673687935, 0.004204560071229935, -0.001134086400270462, 0.0009945333003997803, 0.0010456647723913193, 0.00017123110592365265, 0.0002646711654961109, 0.0008912179619073868 ], "p_value": 0.5343, "mde": 0.0015981726750217166, "mde_rel_pct": 11.09334414171928, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "p": 1.0, "predicted_weight_ratio": 16.0, "observed_weight_ratio": 16.0, "relative_error": 0.0, "confirmed": true } }, "system_verdict": "partial", "practical_verdict": "inconclusive", "mechanism_ok": 1, "system_judged": true }