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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "The idea was tested on registered dynamics with shared rnn_small models and eight paired seeds. It improved mean MSE from 0.0011755236 to 0.0010581258, but p=0.365, so there was no significant win. The mechanism signature was not confirmed because measured information exceeded H_goal and lambda was often inactive.",
  "metrics": {
    "baseline": "MSE mean 0.0011755235827877186; best lr=0.01, pressure=0.2",
    "idea": "MSE mean 0.0010581257738522254; best eta_lambda=0.08, H_goal=0.55; delta=-0.00011739780893549323; p=0.365"
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_dual_dynamics.py",
  "files": [
    "bench_dual_dynamics.py",
    "bench_report.json"
  ],
  "limitations": "Only registered dynamics was tested; no full RSSM, RL return, imagined rollout, Bethe objective, calibration, or FLOP study.",
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.01,
        "pressure": 0.2
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "pressure": 0.0
          },
          "mean": 0.003570652887
        },
        {
          "cfg": {
            "lr": 0.001,
            "pressure": 0.05
          },
          "mean": 0.003419904155
        },
        {
          "cfg": {
            "lr": 0.001,
            "pressure": 0.2
          },
          "mean": 0.003650593804
        },
        {
          "cfg": {
            "lr": 0.003,
            "pressure": 0.0
          },
          "mean": 0.005002945079
        },
        {
          "cfg": {
            "lr": 0.003,
            "pressure": 0.05
          },
          "mean": 0.005080707313
        },
        {
          "cfg": {
            "lr": 0.003,
            "pressure": 0.2
          },
          "mean": 0.005320765835
        },
        {
          "cfg": {
            "lr": 0.01,
            "pressure": 0.0
          },
          "mean": 0.001317500777
        },
        {
          "cfg": {
            "lr": 0.01,
            "pressure": 0.05
          },
          "mean": 0.001284009413
        },
        {
          "cfg": {
            "lr": 0.01,
            "pressure": 0.2
          },
          "mean": 0.001166762275
        }
      ],
      "full": {
        "mean": 0.001175523583,
        "std": 0.000352261609,
        "per_seed": [
          0.00083926064,
          0.00142259663,
          0.000689992623,
          0.001715199207,
          0.001335193519,
          0.001519879559,
          0.001078056986,
          0.000804009498
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.001058125774,
      "std": 0.000258431233,
      "per_seed": [
        0.000787993718,
        0.001001529396,
        0.001303545199,
        0.001240157289,
        0.001351816697,
        0.001315109432,
        0.000720201584,
        0.000744652876
      ],
      "n": 8,
      "best_cfg": {
        "lr": 0.01,
        "pressure": 0.2,
        "eta_lambda": 0.08
      },
      "candidates": [
        {
          "cfg": {
            "lr": 0.01,
            "eta_lambda": 0.04
          },
          "mean": 0.001126833784
        },
        {
          "cfg": {
            "lr": 0.01,
            "eta_lambda": 0.08
          },
          "mean": 0.001058125774
        },
        {
          "cfg": {
            "lr": 0.01,
            "eta_lambda": 0.16
          },
          "mean": 0.001107265515
        }
      ]
    },
    "comparison": {
      "delta_mean": -0.0001173978089,
      "idea_wins": 6,
      "n_pairs": 8,
      "per_seed_diffs": [
        -5.1266921e-05,
        -0.000421067234,
        0.000613552576,
        -0.000475041918,
        1.6623177e-05,
        -0.000204770127,
        -0.000357855402,
        -5.9356622e-05
      ],
      "p_value": 0.365,
      "verdict": "no significant win",
      "system_worked": false
    },
    "mechanism_signature": {
      "H_goal": 0.55,
      "baseline_mean_I_test": 0.867881641,
      "idea_mean_I_test": 0.8641442657,
      "idea_mean_lambda": 0.493634299,
      "confirmed": false
    }
  },
  "system_verdict": "failed",
  "practical_verdict": "inconclusive",
  "mechanism_ok": 0,
  "system_judged": true
}