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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 10,
  "verdict": "The registered dynamics benchmark confirms the safety mechanism but not a task-metric win. Direct interpolation produced 7/9 calibration violations, while the filtered system accepted zero violating steps; nevertheless, filtered topology switching was significantly worse in test MSE, so worked=false.",
  "metrics": {
    "baseline": "Best lr=0.006; 8-seed test MSE mean 0.0007792052.",
    "idea": "Best lr=0.003; 8-seed test MSE mean 0.02425635; paired delta +0.02347715; 0/8 wins."
  },
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.006
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.0015
          },
          "mean": 0.0019174260669387877
        },
        {
          "cfg": {
            "lr": 0.003
          },
          "mean": 0.0009721333481138572
        },
        {
          "cfg": {
            "lr": 0.006
          },
          "mean": 0.0007438920001732185
        }
      ],
      "full": {
        "mean": 0.0007792052056174725,
        "std": 0.00021144133531208117,
        "per_seed": [
          0.0007514121243730187,
          0.0005835454212501645,
          0.0005766521790064871,
          0.0010639582760632038,
          0.0008014270570129156,
          0.0006233751191757619,
          0.0006578490138053894,
          0.001175422454252839
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.024256352335214615,
      "std": 0.0038067872943047523,
      "per_seed": [
        0.021198168396949768,
        0.02363172546029091,
        0.030124139040708542,
        0.02928881347179413,
        0.02132338099181652,
        0.018241815268993378,
        0.024838322773575783,
        0.02540445327758789
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": 0.023477147129597142,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.02044675627257675,
        0.023048180039040744,
        0.029547486861702055,
        0.028224855195730925,
        0.020521953934803605,
        0.017618440149817616,
        0.024180473759770393,
        0.02422903082333505
      ],
      "p_value": 0.0081,
      "mde": 0.0033576493476355656,
      "mde_rel_pct": 430.9069451063066,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "prediction": "direct topology interpolation can transiently violate calibration accuracy; exact receding-horizon filtering accepts no violating step",
      "trained_baseline_unfiltered_violations_of_9": 7,
      "trained_idea_accepted_step_violations": 0,
      "trained_idea_rejected_candidates": 55,
      "trained_idea_transition_steps": 1,
      "predicted_zero_accepted_violations": true,
      "confirmed": true
    },
    "protocol": {
      "paired_seeds": [
        0,
        1,
        2,
        3,
        4,
        5,
        6,
        7
      ],
      "baseline_grid": [
        {
          "lr": 0.0015
        },
        {
          "lr": 0.003
        },
        {
          "lr": 0.006
        }
      ],
      "idea_grid": [
        {
          "lr": 0.0015
        },
        {
          "lr": 0.003
        },
        {
          "lr": 0.006
        }
      ],
      "epochs": 18,
      "n_train": 400,
      "n_test": 200
    }
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 stage2_bench.py",
  "files": [
    "stage2_bench.py",
    "bench_report.json",
    "bench_stdout.txt"
  ],
  "limitations": "Only the registered built-in dynamics track was tested. Activation, spectral norm, latency, memory, hardware speedup, larger models, multiple topology targets, and horizons beyond the implemented single transition were not evaluated. Some runs required the CPU fallback after a shared-GPU cuDNN allocation failure.",
  "system_verdict": "partial",
  "practical_verdict": "harms",
  "mechanism_ok": 1,
  "system_judged": true
}