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

(универсальный)

Ответ агента:

{ "worked": false, "confidence": 10, "verdict": "Bi-Maxwell Muon was evaluated on the registered tabular Friedman#1 track with identical mlp_tiny systems, a tuned baseline sweep, and 8 paired seeds. The trained-model relaxation signature was confirmed, but Bi-Maxwell was significantly worse on test MSE, so there was no practical win.", "metrics": { "baseline": "Muon best lr=0.005, weight_decay=0.0001; mean test MSE 21.5773.", "idea": "Bi-Maxwell best lr=0.003, weight_decay=0.0001, beta_fast=0.90, beta_slow=0.99, weight_fast=0.5; mean test MSE 23.9797.", "comparison": "delta_mean=+2.40236; p_value=0.0081; idea_wins=0/8." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_bench.py", "files": [ "stage2_bench.py", "bench_report.json", "bench_stdout_current.txt", "bi_maxwell_muon.py" ], "limitations": "Only the structurally matched tabular optimizer track was tested; wall-clock and optimizer FLOP overhead were not primary metrics.", "bench_report": { "bench_version": 1, "track": "tabular", "model": "mlp_tiny", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.005, "weight_decay": 0.0001 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 107.3976 }, { "cfg": { "lr": 0.001, "weight_decay": 0.0001 }, "mean": 107.3993 }, { "cfg": { "lr": 0.002, "weight_decay": 0.0 }, "mean": 44.9762 }, { "cfg": { "lr": 0.002, "weight_decay": 0.0001 }, "mean": 44.9787 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 24.1187 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0001 }, "mean": 24.12 }, { "cfg": { "lr": 0.005, "weight_decay": 0.0 }, "mean": 21.4563 }, { "cfg": { "lr": 0.005, "weight_decay": 0.0001 }, "mean": 21.4562 } ], "full": { "mean": 21.5773, "std": 1.5687, "per_seed": [ 22.1565, 22.0342, 22.2544, 19.3798, 19.0824, 22.4242, 21.1202, 24.1668 ], "n": 8 } }, "idea": { "cfg": { "lr": 0.003, "weight_decay": 0.0001, "beta_fast": 0.9, "beta_slow": 0.99, "weight_fast": 0.5 }, "mean": 23.9797, "std": 1.905, "per_seed": [ 25.876, 24.8227, 24.5487, 21.0821, 21.3658, 24.733, 22.761, 26.6481 ], "n": 8 }, "comparison": { "delta_mean": 2.40236, "idea_wins": 0, "n_pairs": 8, "per_seed_diffs": [ 3.7196, 2.7885, 2.2943, 1.7023, 2.2834, 2.3089, 1.6407, 2.4813 ], "p_value": 0.0081, "mde": 0.54695, "mde_rel_pct": 2.5348, "verdict": "idea worse (significant)", "system_worked": false }, "mechanism_signature": { "prediction": "Mixed EMA impulse decay slope lies between fast and slow log-beta values when initialized from a measured trained-model gradient.", "predicted_interval": [ -0.1053605, -0.0100503 ], "observed_mean_slope": -0.0907884, "confirmed": true }, "custom_track": null }, "system_verdict": "partial", "practical_verdict": "harms", "mechanism_ok": 1, "system_judged": true }