Invariant nonstandard residual blocks / report_bench_2026-09-04T200606.md

✓✓ Beats tuned baseline

Raw ⬇ ZIP

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

(универсальный)

Ответ агента:

{ "worked": true, "confidence": 8, "verdict": "Implemented a matched dynamics-track residual RNN benchmark in stage2_bench.py using the canonical bench trainer, eight paired seeds, baseline learning-rate sweep, parity-complete idea sweep, and a trained-model mechanism signature. The denominator-damped update significantly improved test MSE: 0.002001 versus 0.061238 for the tuned baseline, with paired delta -0.059237 and permutation p=0.0081; all eight pairs favored the idea. The trained-state update proxy was 4.464 versus 21.130, ratio 0.211, confirming the predicted damping effect.", "metrics": { "baseline": "Dynamics test MSE mean 0.0612377168, std 0.0272115491, best lr=0.003; sweep means lr 0.001=0.0826119585, lr 0.003=0.0658643981, lr 0.01=0.1500396333.", "idea": "Dynamics test MSE mean 0.0020011790, std 0.0007409601, best lr=0.001; sweep means lr 0.001=0.0020011790, lr 0.003=0.0025825649, lr 0.01=0.0343854544. Paired delta=-0.0592365378, p=0.0081, 8/8 wins. Observed update-proxy ratio=0.2113 versus predicted q/h=0.3333; confirmed=true." }, "bench_report": { "bench_version": 1, "track": "dynamics", "model": "residual_rnn_small", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.003 }, "sweep": [ { "cfg": { "lr": 0.001 }, "mean": 0.08261195849627256 }, { "cfg": { "lr": 0.003 }, "mean": 0.06586439814418554 }, { "cfg": { "lr": 0.01 }, "mean": 0.15003963327035308 } ], "full": { "mean": 0.061237716814503074, "std": 0.027211549075682028, "per_seed": [ 0.04314250871539116, 0.09609165042638779, 0.08621135354042053, 0.038012079894542694, 0.07434208691120148, 0.028976354748010635, 0.09323041886091232, 0.029895281419157982 ], "n": 8 } }, "idea": { "mean": 0.0020011790038552135, "std": 0.0007409600802683955, "per_seed": [ 0.0010639947140589356, 0.0032321331091225147, 0.0009753473568707705, 0.0024151953402906656, 0.002823632676154375, 0.0017429819563403726, 0.0018509947694838047, 0.0019051521085202694 ], "n": 8 }, "comparison": { "delta_mean": -0.05923653781064786, "idea_wins": 8, "n_pairs": 8, "per_seed_diffs": [ -0.04207851400133222, -0.09285951731726527, -0.08523600618354976, -0.03559688455425203, -0.0715184542350471, -0.027233372791670263, -0.09137942409142852, -0.027990129310637712 ], "p_value": 0.0081, "mde": 0.02416027485033725, "mde_rel_pct": 39.453258722107215, "verdict": "idea better (significant)", "system_worked": true }, "math_check": { "h": [ 0.5, 1.0, 2.0, 4.0, 8.0, 32.0 ], "q": [ 0.4, 0.6666666666666666, 1.0, 1.3333333333333333, 1.6, 1.8823529411764706 ], "q_bound": 2.0, "euler_nonexpansive_test": false, "damped_nonexpansive_test": true }, "mechanism_signature": { "prediction": "denominator block has smaller trained-state update proxy at h=4", "nominal_h": 4.0, "alpha": 0.5, "observed_baseline_update": 21.13020896911621, "observed_idea_update": 4.463923364877701, "observed_update_ratio": 0.21125789013266005, "predicted_upper_ratio_from_q_over_h": 0.3333333333333333, "confirmed": true } }, "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_bench.py", "files": [ "stage2_bench.py", "bench_report.json" ], "limitations": "Only the structurally matched built-in dynamics track was tested. The implementation uses a two-stage explicit midpoint-style block with denominator damping rather than a fully implicit general-linear solve; wall-clock/FLOP efficiency, CIFAR-10, sequence, and long-horizon generalization were not evaluated.", "system_verdict": "worked", "practical_verdict": "helps", "mechanism_ok": 1, "system_judged": true }