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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 10,
  "verdict": "The dissipative action-coordinate parameterization was implemented with the shared rnn_small GRU core and evaluated on the structurally matched registered dynamics track. The trained-model mechanism signature confirmed the predicted contraction quantitatively, but the idea's test MSE was dramatically worse than the tuned baseline and the paired permutation test found a significant disadvantage.",
  "metrics": {
    "baseline": "Best lr=0.01, weight_decay=0.0; 8-seed test MSE mean=3.3532741667841037e-06, std=1.1020743584708325e-06.",
    "idea": "Best lr=0.01, weight_decay=0.0, mu=0.3, Lf=0.2; 8-seed test MSE mean=0.5757362097501755, std=0.020123708021359012; paired delta=+0.5757328564760087, p=0.0081; idea worse significantly."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_experiment.py",
  "files": [
    "bench_experiment.py",
    "bench_report.json",
    "rerun_stdout.json"
  ],
  "limitations": "Only the registered built-in dynamics track was tested. Diffusion-policy sampling, multimodal action coverage, sample-size scaling, and Euler-step sensitivity were not tested. The built-in target is future pendulum angle rather than direct vector-field prediction, so this is a minimal transfer of the proposed action-coordinate drift structure.",
  "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": 3.534919142111903e-05
        },
        {
          "cfg": {
            "lr": 0.001,
            "weight_decay": 0.0001
          },
          "mean": 0.00018523921971791424
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0
          },
          "mean": 7.330269568228687e-06
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0001
          },
          "mean": 6.71145257911121e-05
        },
        {
          "cfg": {
            "lr": 0.01,
            "weight_decay": 0.0
          },
          "mean": 3.9207230884130695e-06
        },
        {
          "cfg": {
            "lr": 0.01,
            "weight_decay": 0.0001
          },
          "mean": 2.6831472723642946e-05
        }
      ],
      "full": {
        "mean": 3.3532741667841037e-06,
        "std": 1.1020743584708325e-06,
        "per_seed": [
          4.035734946228331e-06,
          3.1895601750875358e-06,
          2.5228919184883125e-06,
          5.934705313848099e-06,
          3.3925441584869986e-06,
          2.830740641002194e-06,
          2.4708413093321724e-06,
          2.449174871799187e-06
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.5757362097501755,
      "std": 0.020123708021359012,
      "per_seed": [
        0.5845423936843872,
        0.5802162289619446,
        0.5462422966957092,
        0.6178359985351562,
        0.5681619644165039,
        0.5554149746894836,
        0.5804445147514343,
        0.5730313062667847
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": 0.5757328564760087,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.584538357949441,
        0.5802130394017695,
        0.5462397738037907,
        0.6178300638298424,
        0.5681585718723454,
        0.5554121439488426,
        0.580442043910125,
        0.5730288570919129
      ],
      "p_value": 0.0081,
      "mde": 0.017987483920840156,
      "mde_rel_pct": 536415.5457079945,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "mu": 0.3,
      "Lf": 0.2,
      "predicted_bound": -0.09999999999999998,
      "idea_observed_residual_fd_lipschitz_max": 0.1959163397550583,
      "idea_observed_one_sided_ratio_max": -0.10408368706703186,
      "baseline_observed_one_sided_ratio_max": 0.009726762771606445,
      "confirmed": true
    },
    "math_sanity_check": {
      "mu": 0.3,
      "Lf": 0.2,
      "theoretical_bound": -0.09999999999999998,
      "max_residual_fd_lipschitz": 0.1999997138561903,
      "max_one_sided_ratio": -0.09999965212070401,
      "violation_fraction": 5e-06,
      "confirmed": false
    },
    "idea_grid": [
      {
        "cfg": {
          "lr": 0.01,
          "weight_decay": 0.0
        },
        "mean": 0.5757362097501755
      },
      {
        "cfg": {
          "lr": 0.003,
          "weight_decay": 0.0
        },
        "mean": 0.575765073299408
      },
      {
        "cfg": {
          "lr": 0.001,
          "weight_decay": 0.0
        },
        "mean": 0.5757729709148407
      }
    ],
    "idea_best_cfg": {
      "lr": 0.01,
      "weight_decay": 0.0
    },
    "custom_track": null
  },
  "system_verdict": "partial",
  "practical_verdict": "harms",
  "mechanism_ok": 1,
  "system_judged": true
}