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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "Implemented a custom SPD covariance-classification track because no built-in track contains the required SPD covariance structure. The trained Cholesky-structured system preserved positive-definite prototype factors and finite logits, confirming the stability mechanism, but its mean test error was 0.25125 versus 0.005625 for the baseline; paired delta was +0.245625 with permutation p=0.2308, so there was no win.",
  "metrics": {
    "baseline": "Mean test classification error 0.005625 across 8 seeds; best baseline sweep setting lr=0.001, epochs=18.",
    "idea": "Mean test classification error 0.251250 across 8 seeds; selected idea setting lr=0.003, epochs=18; observed prototype minimum eigenvalue 0.420817 and finite-logit fraction 1.0."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 spd_bench.py",
  "files": [
    "spd_bench.py",
    "bench_report.json"
  ],
  "limitations": "The custom track is synthetic rather than a real-world covariance dataset; matrix dimension was fixed at 5, M was fixed to the identity, and the matrix-power implementation used symmetric eigendecomposition rather than a genuinely Cholesky-only operator. Wall-clock speed, peak memory, and dimensions up to 128 were not benchmarked. The custom track registration into the shared read-only bench was not performed by this experiment.",
  "bench_report": {
    "bench_version": 1,
    "track": "spd_covariance_classification",
    "model": "mlp_tiny",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.001,
        "epochs": 18
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "epochs": 18
          },
          "mean": 0.004999999888241291
        },
        {
          "cfg": {
            "lr": 0.003,
            "epochs": 18
          },
          "mean": 0.006249999860301614
        },
        {
          "cfg": {
            "lr": 0.01,
            "epochs": 18
          },
          "mean": 0.007499999832361937
        }
      ],
      "full": {
        "mean": 0.005624999874271452,
        "std": 0.0029973946350734574,
        "per_seed": [
          0.004999999888241291,
          0.009999999776482582,
          0.004999999888241291,
          0.0,
          0.009999999776482582,
          0.004999999888241291,
          0.004999999888241291,
          0.004999999888241291
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.25124998565297574,
      "std": 0.32049520383799557,
      "per_seed": [
        0.004999999888241291,
        0.004999999888241291,
        0.6649999618530273,
        0.004999999888241291,
        0.0,
        0.6649999618530273,
        0.6649999618530273,
        0.0
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": 0.24562498577870429,
      "idea_wins": 3,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.0,
        -0.004999999888241291,
        0.659999961964786,
        0.004999999888241291,
        -0.009999999776482582,
        0.659999961964786,
        0.659999961964786,
        -0.004999999888241291
      ],
      "p_value": 0.2308,
      "mde": 0.28693631664685193,
      "mde_rel_pct": 5101.090187740063,
      "verdict": "no significant win",
      "system_worked": false
    },
    "mechanism_signature": {
      "prediction": "factor-generated prototypes remain SPD and logits finite after NN training",
      "predicted_min_eigenvalue_bound": 0.0,
      "observed_min_prototype_eigenvalue": 0.42081671953201294,
      "observed_finite_logit_fraction": 1.0,
      "confirmed": true,
      "idea_settings_tried": [
        {
          "lr": 0.001,
          "epochs": 18
        },
        {
          "lr": 0.003,
          "epochs": 18
        },
        {
          "lr": 0.01,
          "epochs": 18
        }
      ]
    }
  },
  "system_verdict": "partial",
  "practical_verdict": "inconclusive",
  "mechanism_ok": 1,
  "system_judged": true
}