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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "The registered dynamics track compared matched GRU systems differing only in hard versus Filippov routing. Filippov routing reduced mean test MSE from 5.99486e-05 to 5.37679e-05, but p=0.3579 and therefore there was no significant win. The trained-model mechanism signature was unconfirmed because no qualifying attracting sliding examples occurred.",
  "metrics": {
    "baseline": "best lr=0.04; mean MSE=5.994863204250578e-05; std=2.276605305533657e-05",
    "idea": "lr=0.04, epsilon=0.08, tau=0.016; mean MSE=5.376790909394913e-05; std=1.6198896783739743e-05"
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_bench.py",
  "files": [
    "filippov_layer.py",
    "check_filippov.py",
    "stage2_bench.py",
    "bench_report.json"
  ],
  "limitations": "Only dynamics was tested, with 12 epochs and 1200/300 train/test examples. No trained-model sliding cases were observed, so the quantitative mechanism prediction was not confirmed.",
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.04
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001
          },
          "mean": 0.001978325890377164
        },
        {
          "cfg": {
            "lr": 0.003
          },
          "mean": 0.0007930935244075954
        },
        {
          "cfg": {
            "lr": 0.01
          },
          "mean": 0.00026890657863987144
        },
        {
          "cfg": {
            "lr": 0.02
          },
          "mean": 0.00015998855269572232
        },
        {
          "cfg": {
            "lr": 0.04
          },
          "mean": 5.8887695558951236e-05
        },
        {
          "cfg": {
            "lr": 0.08
          },
          "mean": 0.0002853818004950881
        }
      ],
      "full": {
        "mean": 5.994863204250578e-05,
        "std": 2.276605305533657e-05,
        "per_seed": [
          7.932300650281832e-05,
          3.942438343074173e-05,
          6.531396502396092e-05,
          5.1489427278283983e-05,
          0.00010841330367838964,
          4.656820237869397e-05,
          3.343872958794236e-05,
          5.561803845921531e-05
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 5.376790909394913e-05,
      "std": 1.6198896783739743e-05,
      "per_seed": [
        7.933847518870607e-05,
        2.778465386654716e-05,
        6.544616189785302e-05,
        5.777328260592185e-05,
        6.612198922084644e-05,
        4.65624216303695e-05,
        3.343872958794236e-05,
        5.367755875340663e-05
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": -6.1807229485566495e-06,
      "idea_wins": 4,
      "n_pairs": 8,
      "p_value": 0.3579,
      "verdict": "no significant win",
      "system_worked": false
    },
    "mechanism_signature": {
      "prediction": "During attracting sliding, predicted normal velocity ratio <0.05 of typical vector-field norm.",
      "trained_model_measurements": [],
      "mean_ratio": null,
      "confirmed": false
    }
  },
  "system_verdict": "failed",
  "practical_verdict": "inconclusive",
  "mechanism_ok": 0,
  "system_judged": true
}