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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "Implemented and evaluated a custom complex-array channel regression track with matched MLP encoders and an analytic rank-one steering-atom decoder. The idea's best result was MSE 0.0076194 versus baseline 0.0067185; paired delta was +0.0009009 and the permutation p-value was 0.0081, yielding the significant-worse verdict. The analytic decoder therefore did not produce a neural-training win in this benchmark.",
  "metrics": {
    "baseline": "Tuned learned-output MLP, lr=0.01: full 8-seed MSE 0.0067185 ± 0.000713",
    "idea": "Analytic K=2 steering-atom MLP, lr=0.01: full 8-seed MSE 0.0076194 ± 0.000828; paired delta +0.0009009, p=0.0081"
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 run_bench.py",
  "files": [
    "array_track.py",
    "run_bench.py",
    "bench_report.json"
  ],
  "limitations": "The custom track uses 8x8 synthetic two-path channels, not a real wireless dataset or 32x32/48x48 transfer. It does not test quantization, variable active-slot masking, rate-vs-error curves, decoder FLOPs, or learned off-grid offsets/Taylor decoding; K was fixed at 2. The reported Taylor signature is an analytic steering-vector probe alongside a trained-model output RMS, not a robust learned-coordinate behavioral test, so it should not be treated as independent mechanism confirmation.",
  "bench_report": {
    "bench_version": 1,
    "track": "complex_array_channel",
    "model": "mlp_tiny",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.01,
        "k": 2
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "k": 2
          },
          "mean": 0.007402819930575788
        },
        {
          "cfg": {
            "lr": 0.003,
            "k": 2
          },
          "mean": 0.0065840884344652295
        },
        {
          "cfg": {
            "lr": 0.01,
            "k": 2
          },
          "mean": 0.006486733560450375
        }
      ],
      "full": {
        "mean": 0.006718522927258164,
        "std": 0.0007129592821712252,
        "per_seed": [
          0.006483251228928566,
          0.00544784078374505,
          0.007208648603409529,
          0.006807193625718355,
          0.007167691830545664,
          0.007836345583200455,
          0.005902399308979511,
          0.006894812453538179
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.007619430776685476,
      "std": 0.0008275326682381549,
      "per_seed": [
        0.0074270921759307384,
        0.006459011230617762,
        0.008539552800357342,
        0.007304489612579346,
        0.007882530800998211,
        0.00901730451732874,
        0.006567466538399458,
        0.007757998537272215
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": 0.0009009078494273126,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.0009438409470021725,
        0.0010111704468727112,
        0.001330904196947813,
        0.0004972959868609905,
        0.0007148389704525471,
        0.0011809589341282845,
        0.0006650672294199467,
        0.0008631860837340355
      ],
      "p_value": 0.0081,
      "mde": 0.0002304888187870338,
      "mde_rel_pct": 3.43064720151363,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "prediction": "Taylor steering error scales quadratically in offset",
      "delta": 0.001,
      "predicted_error_order": 2.0,
      "observed_error_over_delta_squared": 34.418742870911956,
      "trained_model_output_rms": 0.011084591038525105,
      "confirmed": true
    },
    "idea_sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "k": 2
        },
        "mean": 0.007869121560361236,
        "std": 0.0008392636492011609
      },
      {
        "cfg": {
          "lr": 0.003,
          "k": 2
        },
        "mean": 0.007681989867705852,
        "std": 0.0008145156626225875
      },
      {
        "cfg": {
          "lr": 0.01,
          "k": 2
        },
        "mean": 0.007619430776685476,
        "std": 0.0008275326682381549
      }
    ],
    "custom_track": {
      "name": "complex_array_channel",
      "file": "array_track.py",
      "domain": "array-valued complex low-rank tensors"
    }
  },
  "system_verdict": "partial",
  "practical_verdict": "harms",
  "mechanism_ok": 1,
  "system_judged": true
}