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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "Built and evaluated a matched residual MLP on the Friedman#1 tabular benchmark using the canonical bench training loop, an 8-seed paired comparison, and a parity-preserving baseline sweep. The mathematical scaling check passed (theory H=0.8, observed H=0.821, scaled RMS ratio=1.063), but the idea had higher test MSE than the tuned iid baseline (7.037 vs 6.537; paired delta +0.499, p=0.0081) and lost on all 8 seeds. Therefore the stability signal does not translate into a benchmark training improvement in this setup.",
  "metrics": {
    "baseline": "Friedman#1 tabular residual MLP, iid gates, best swept config lr=0.006, exponent=0.5, 18 epochs; full 8-seed mean test MSE=6.5375, std=0.7963.",
    "idea": "Same residual MLP/data/Adam/batch/epochs and lr union, correlated Gaussian gates with alpha=0.4 and best idea exponent=0.7; full 8-seed mean test MSE=7.0365, std=0.6551.",
    "paired_delta": "idea - baseline = +0.4990 MSE; idea wins 0/8 seeds; permutation p=0.0081; verdict=idea worse (significant).",
    "math_check": "H theory=0.8, observed H=0.8213, scaled-sum RMS values=[1.3915,1.2675,1.4158,1.2996,1.4795], last/first=1.0632, passed=true.",
    "mechanism_signature": "Predicted adjacent gate correlation=0.7579, observed mean across trained benchmark systems=0.0362, confirmed=false."
  },
  "bench_report": {
    "bench_version": 1,
    "track": "tabular",
    "model": "local_residual_mlp",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.006,
        "epochs": 18,
        "exponent": 0.5
      },
      "full": {
        "mean": 6.537471890449524,
        "std": 0.7962674958223888,
        "n": 8,
        "per_seed": [
          5.471584320068359,
          6.244831562042236,
          7.435494422912598,
          5.489790439605713,
          7.0974907875061035,
          5.940451622009277,
          7.106790065765381,
          7.513341903686523
        ]
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "epochs": 18,
            "exponent": 0.5
          },
          "mean": 7.946774959564209
        },
        {
          "cfg": {
            "lr": 0.001,
            "epochs": 18,
            "exponent": 0.8
          },
          "mean": 10.225590229034424
        },
        {
          "cfg": {
            "lr": 0.003,
            "epochs": 18,
            "exponent": 0.5
          },
          "mean": 7.576857328414917
        },
        {
          "cfg": {
            "lr": 0.003,
            "epochs": 18,
            "exponent": 0.8
          },
          "mean": 7.316884994506836
        },
        {
          "cfg": {
            "lr": 0.006,
            "epochs": 18,
            "exponent": 0.5
          },
          "mean": 6.160425186157227
        },
        {
          "cfg": {
            "lr": 0.006,
            "epochs": 18,
            "exponent": 0.8
          },
          "mean": 7.347582578659058
        }
      ]
    },
    "idea": {
      "cfg": {
        "lr": 0.006,
        "epochs": 18,
        "exponent": 0.7
      },
      "mean": 7.036507606506348,
      "std": 0.6550880229661826,
      "n": 8,
      "per_seed": [
        7.978086948394775,
        6.377083778381348,
        7.57796049118042,
        5.985846996307373,
        7.230166435241699,
        6.412862300872803,
        7.17480993270874,
        7.555243968963623
      ]
    },
    "comparison": {
      "delta_mean": 0.49903571605682373,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        2.506502628326416,
        0.13225221633911133,
        0.14246606826782227,
        0.49605655670166016,
        0.1326756477355957,
        0.4724106788635254,
        0.06801986694335938,
        0.04190206527709961
      ],
      "p_value": 0.0081,
      "mde": 0.6939341374663044,
      "mde_rel_pct": 10.61471695931971,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "predicted_adjacent_corr": 0.757858283255199,
      "observed_adjacent_corr_mean": 0.03618130783796214,
      "confirmed": false
    },
    "math_check": {
      "theory_H": 0.8,
      "observed_H": 0.8212929503641088,
      "rms_ratio_last_first": 1.0632182958840077,
      "passed": true
    }
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_experiment.py",
  "files": [
    "bench_experiment.py",
    "bench_report.json"
  ],
  "limitations": "Only the mandatory structurally appropriate tabular track was tested; no CIFAR-10, sequence, dynamics, q>1 neural-network variant, deeper residual networks, or longer training budgets were evaluated. The local implementation uses a small depth-8 residual MLP and finite-length Cholesky correlation generation rather than FFT generation or large-depth asymptotics.",
  "system_verdict": "failed",
  "practical_verdict": "harms",
  "mechanism_ok": 0,
  "system_judged": true
}