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

(универсальный)

Ответ агента:

{
  "worked": true,
  "confidence": 9,
  "verdict": "Implemented small-gain recurrent operator projection on the registered dynamics track using matched rnn_small GRU systems. The idea achieved significantly lower test MSE than the tuned baseline on all 8 paired seeds, with permutation p=0.0081. The trained-model gain proxy stayed below the 0.90 cap, but this is evidence from one small benchmark only.",
  "metrics": {
    "baseline": "Unconstrained rnn_small, tuned lr=0.005: mean test MSE 0.0003818532, std 0.0001613648.",
    "idea": "Projected rnn_small, gain cap=0.90 and lr=0.005: mean test MSE 0.0001617695, std 0.0000254929; paired delta -0.0002200837; 8/8 wins; p=0.0081."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_experiment.py",
  "files": [
    "bench_experiment.py",
    "bench_report.json"
  ],
  "limitations": "Only the registered built-in dynamics track was tested. No larger model, alternative caps, explicit complete nonlinear GRU gain matrix, long-horizon rollout metric, FLOP-normalized comparison, or repeated rerun was tested; recurrent gate operator norms are an engineering proxy for the full small-gain certificate.",
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.005,
        "rho_cap": null
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "rho_cap": null
          },
          "mean": 0.0012295851483941078
        },
        {
          "cfg": {
            "lr": 0.003,
            "rho_cap": null
          },
          "mean": 0.0006583010253962129
        },
        {
          "cfg": {
            "lr": 0.005,
            "rho_cap": null
          },
          "mean": 0.0004532121238298714
        }
      ],
      "full": {
        "mean": 0.00038185320590855554,
        "std": 0.00016136478368296086,
        "per_seed": [
          0.00046871721860952675,
          0.0002616161364130676,
          0.00031301137642003596,
          0.0007695037638768554,
          0.0003513892588671297,
          0.0002713301219046116,
          0.0002475073270034045,
          0.00037175044417381287
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.00016176951703528175,
      "std": 2.54928932189329e-05,
      "per_seed": [
        0.00013927802501711994,
        0.0001758444996085018,
        0.0001558296353323385,
        0.00020030193263664842,
        0.0001420884218532592,
        0.00012167391105322167,
        0.00016640945977997035,
        0.00019273025100119412
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": -0.00022008368887327379,
      "idea_wins": 8,
      "n_pairs": 8,
      "per_seed_diffs": [
        -0.0003294391935924068,
        -8.577163680456579e-05,
        -0.00015718174108769745,
        -0.000569201831240207,
        -0.00020930083701387048,
        -0.00014965621085138991,
        -8.109786722343415e-05,
        -0.00017902019317261875
      ],
      "p_value": 0.0081,
      "mde": 0.0001347953881261106,
      "mde_rel_pct": 35.3003159435018,
      "verdict": "idea better (significant)",
      "system_worked": true
    },
    "mechanism_signature": {
      "prediction": "Projected recurrent operator-gain proxy stays <= 0.90.",
      "predicted_max": 0.7689717411994934,
      "observed": [
        {
          "seed": 0,
          "predicted_gain_proxy": 0.7254310846328735,
          "observed_output_gain": 0.0707114040851593
        },
        {
          "seed": 1,
          "predicted_gain_proxy": 0.7678388953208923,
          "observed_output_gain": 0.09257301688194275
        },
        {
          "seed": 2,
          "predicted_gain_proxy": 0.7689717411994934,
          "observed_output_gain": 0.07351840287446976
        },
        {
          "seed": 3,
          "predicted_gain_proxy": 0.7672179341316223,
          "observed_output_gain": 0.07399722933769226
        }
      ],
      "confirmed": true
    },
    "idea_sweep": {
      "0.001": {
        "mean": 0.00032572537929809187,
        "std": 6.515504960120716e-05
      },
      "0.003": {
        "mean": 0.0002240399771835655,
        "std": 3.985152266293653e-05
      },
      "0.005": {
        "mean": 0.00016176951703528175,
        "std": 2.54928932189329e-05
      }
    }
  },
  "system_verdict": "worked",
  "practical_verdict": "helps",
  "mechanism_ok": 1,
  "system_judged": true
}