Finite-Excitation Latent Replay / report_bench_2026-09-01T235855.md

Failed on benchmark

Raw ⬇ ZIP

Стенд-проверка (stage-2) · промт оператора:

(универсальный)

Ответ агента:

{ "worked": false, "confidence": 10, "verdict": "Implemented Finite-Excitation Latent Replay on the structurally matched controlled-pendulum dynamics track using the shared rnn_small architecture. The perturbation-corrected excitation certificate never became positive, so the gated system performed no parameter updates and was dramatically worse than Adam. The paired difference was +0.7322945 MSE with permutation p=0.0081; the mechanism signature was not confirmed.", "metrics": { "baseline": "Adam sweep best lr=0.01, weight_decay=0.0; 8-seed test MSE mean=0.000239025, std=0.000163569.", "idea": "Integral-regression excitation-gated replay, best tested lr=0.001/0.003/0.01 with gamma=0, eps=0.035, replay=4; 8-seed test MSE mean=0.732534, std=0.076278.", "paired_delta": "idea - baseline = +0.7322945 MSE; 0/8 idea wins; permutation p=0.0081; verdict=idea worse (significant)." }, "bench_report": { "bench_version": 1, "track": "dynamics", "model": "rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 0.0012379241 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001 }, "mean": 0.0012265432 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.0008770637 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 0.0008647405 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "mean": 0.0003551548 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0001 }, "mean": 0.0004823483 } ], "full": { "mean": 0.0002390248, "std": 0.0001635689, "per_seed": [ 0.0003642804, 0.0005520924, 0.0003466835, 0.000157563, 0.0002606253, 0.0001060843, 3.41185e-05, 9.07508e-05 ], "n": 8 } }, "idea": { "mean": 0.732533522, "std": 0.0762778934, "per_seed": [ 0.6195922494, 0.826359868, 0.7263849378, 0.8282962441, 0.6348818541, 0.7560707927, 0.6810366511, 0.7876455784 ], "n": 8 }, "comparison": { "delta_mean": 0.7322944972, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 0.619228, 0.8258078, 0.7260383, 0.8281387, 0.6346212, 0.7559647, 0.6810025, 0.7875548 ], "p_value": 0.0081, "mde": 0.06817817, "mde_rel_pct": 28523.48, "verdict": "idea worse (significant)", "system_worked": false }, "mechanism_signature": { "prediction": "parameter updates become active after q exceeds gamma and loss then decreases", "predicted_vs_observed": { "predicted_positive_q_activation": true, "observed_positive_q_rate_mean": 0.0, "observed_activation_rate_mean": 0.0, "observed_post_activation_loss_drop_mean": 0.0, "activation_epoch_values": [ null, null, null, null, null, null, null, null ] }, "confirmed": false }, "idea_sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0, "gamma": 0.0, "eps": 0.035, "replay": 4 }, "mean": 0.7501583248 } ], "protocol": { "seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "n_train": 900, "n_test": 300, "epochs": 14, "structural_match": "controlled pendulum rollout / latent dynamics", "custom_track": null } }, "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_felr.py", "files": [ "bench_felr.py", "bench_report.json" ], "limitations": "Only the built-in controlled-pendulum dynamics track was tested. The intervention used a fixed hand-designed feature map and empirical constant uncertainty rather than a learned encoder ensemble; no Lorenz, latent ODE, longer horizons, or alternative uncertainty calibration was evaluated. The benchmark harness path was available at /home/maxwelhelp/all/math2nn/bench despite the initially stated absolute location.", "system_verdict": "failed", "practical_verdict": "harms", "mechanism_ok": 0, "system_judged": true }