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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "Resolution-Gated Dual Masking was implemented and evaluated on the registered Friedman#1 tabular track with the matched mlp_tiny architecture and identical train_model path. The idea obtained lower mean MSE (7.01413 vs 7.05324), but the paired permutation test was non-significant (delta=-0.03911, p=0.50145), so this is not a demonstrated win. The trained-model mechanism signature confirmed kappa=1.0 and fallback away from the entropy mask on all 8 seeds.",
  "metrics": {
    "baseline": "Best tuned baseline MSE=7.0532397628, lr=0.006, weight_decay=0.0001, epochs=12.",
    "idea": "Best gated-mask MSE=7.0141316056, paired delta=-0.0391081572, 4/8 wins, permutation p=0.50145."
  },
  "bench_report": {
    "bench_version": 1,
    "track": "tabular",
    "model": "mlp_tiny",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.006,
        "weight_decay": 0.0001,
        "epochs": 12
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.0015,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 13.675705671310425
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 9.352656126022339
        },
        {
          "cfg": {
            "lr": 0.006,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 7.226524114608765
        },
        {
          "cfg": {
            "lr": 0.0015,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 13.609990000724792
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 9.076743841171265
        },
        {
          "cfg": {
            "lr": 0.006,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 7.0532397627830505
        }
      ],
      "full": {
        "mean": 7.0532397627830505,
        "std": 0.2264412705403704,
        "per_seed": [
          6.572167873382568,
          7.1756792068481445,
          7.042057037353516,
          7.032951831817627,
          7.018033504486084,
          7.022153854370117,
          7.112776756286621,
          7.450098037719727
        ],
        "n": 8
      }
    },
    "idea": {
      "best_cfg": {
        "lr": 0.006,
        "weight_decay": 0.0001,
        "epochs": 12
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.0015,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 13.917831420898438
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 9.332432746887207
        },
        {
          "cfg": {
            "lr": 0.006,
            "weight_decay": 0.0,
            "epochs": 12
          },
          "mean": 7.183628082275391
        },
        {
          "cfg": {
            "lr": 0.0015,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 13.897738456726074
        },
        {
          "cfg": {
            "lr": 0.003,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 9.150044441223145
        },
        {
          "cfg": {
            "lr": 0.006,
            "weight_decay": 0.0001,
            "epochs": 12
          },
          "mean": 7.014131605625153
        }
      ],
      "full": {
        "mean": 7.014131605625153,
        "std": 0.24308566746895208,
        "per_seed": [
          6.6641693115234375,
          7.260889053344727,
          6.737545013427734,
          6.856451034545898,
          6.927297592163086,
          7.039077281951904,
          7.303320407867432,
          7.324303150177002
        ],
        "n": 8
      }
    },
    "comparison": {
      "delta_mean": -0.03910815715789795,
      "idea_wins": 4,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.09200143814086914,
        0.08520984649658203,
        -0.30451202392578125,
        -0.17650079727172852,
        -0.09073591232299805,
        0.01692342758178711,
        0.19054365158081055,
        -0.1257948875427246
      ],
      "p_value": 0.50145,
      "mde": 0.13703746835599878,
      "mde_rel_pct": 1.9429010350546605,
      "verdict": "no measurable effect",
      "system_worked": false
    },
    "mechanism_signature": {
      "prediction": "kappa>=0.90 routes away from entropy mask and avoids materially worse trained-model MSE",
      "observed": [
        {
          "seed": 0,
          "kappa": 1.0,
          "gamma": 13.140283130679085,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": 0.09200143814086914
        },
        {
          "seed": 1,
          "kappa": 1.0,
          "gamma": 16.30871552694337,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": 0.08520984649658203
        },
        {
          "seed": 2,
          "kappa": 1.0,
          "gamma": 12.264759413539208,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": -0.30451202392578125
        },
        {
          "seed": 3,
          "kappa": 1.0,
          "gamma": 13.648493168331017,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": -0.17650079727172852
        },
        {
          "seed": 4,
          "kappa": 1.0,
          "gamma": 15.0961523290884,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": -0.09073591232299805
        },
        {
          "seed": 5,
          "kappa": 1.0,
          "gamma": 11.799654276914342,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": 0.01692342758178711
        },
        {
          "seed": 6,
          "kappa": 1.0,
          "gamma": 15.35830432998792,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": 0.19054365158081055
        },
        {
          "seed": 7,
          "kappa": 1.0,
          "gamma": 13.753987138531764,
          "used_entropy": false,
          "predicted_fallback": true,
          "observed_subset_minus_full_mse": -0.1257948875427246
        }
      ],
      "high_kappa_n": 8,
      "high_kappa_observed_mean_delta": -0.03910815715789795,
      "confirmed": true
    }
  },
  "how_to_run": "cd /home/maxwelhelp/all/math2nn/experiments/exp3124_resolution-gated-dual-masking && /home/maxwelhelp/main/bin/python3 stage2_bench.py",
  "files": [
    "stage2_bench.py",
    "bench_report.json"
  ],
  "limitations": "Only the structurally matched registered tabular track was tested. Learned mixture-of-experts gating, broader threshold sweeps, speed/FLOP measurements, and larger datasets were not tested.",
  "system_verdict": "partial",
  "practical_verdict": "no_effect",
  "mechanism_ok": 1,
  "system_judged": true
}