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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 9,
  "verdict": "Built a local Stieltjes event-driven residual state model and compared it end-to-end against an explicit residual Euler baseline on the structurally matched sequence forecasting track. The mechanism signature confirmed measured zero state change on clock-inactive intervals, but the idea was significantly worse: test MSE 0.70479 versus 0.17484, paired delta +0.52995, permutation p=0.0081. Therefore the stage-2 neural-training transfer did not provide a win.",
  "metrics": {
    "baseline": "Sequence forecasting, explicit residual Euler, best lr=0.003; 8-seed test MSE mean 0.1748437565, std 0.0253561090.",
    "idea": "Sequence forecasting, implicit Stieltjes residual with event clock, best lr=0.01; 8-seed test MSE mean 0.7047909945, std 0.0557582447.",
    "paired_delta": "idea - baseline = +0.5299472380 MSE; permutation p=0.0081; verdict idea worse (significant).",
    "mechanism_signature": {
      "predicted_inactive_state_change": 0.0,
      "observed_baseline_inactive_change": 0.0,
      "observed_idea_inactive_change": 0.0,
      "observed_baseline_event_change": 3.5917565823,
      "observed_idea_event_change": 1.8746105433,
      "idea_active_fraction": 0.9478124976,
      "confirmed": true
    }
  },
  "bench_report": {
    "bench_version": 1,
    "track": "sequence",
    "model": "custom_residual_state",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.003,
        "clock_scale": 1.0
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "clock_scale": 1.0
          },
          "mean": 0.2847412378
        },
        {
          "cfg": {
            "lr": 0.003,
            "clock_scale": 1.0
          },
          "mean": 0.1826061681
        },
        {
          "cfg": {
            "lr": 0.01,
            "clock_scale": 1.0
          },
          "mean": 0.2004329786
        }
      ],
      "full": {
        "mean": 0.1748437565,
        "std": 0.025356109,
        "per_seed": [
          0.1715032756,
          0.1548469514,
          0.1772066206,
          0.2268678248,
          0.1976654828,
          0.1399528235,
          0.172495082,
          0.1582119912
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.7047909945,
      "std": 0.0557582447,
      "per_seed": [
        0.7154005766,
        0.6783341765,
        0.7505328059,
        0.6759172082,
        0.5878732204,
        0.7312020659,
        0.786696732,
        0.7123711705
      ],
      "n": 8,
      "best_cfg": {
        "lr": 0.01,
        "clock_scale": 1.0
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "clock_scale": 1.0
          },
          "mean": 0.7530882806
        },
        {
          "cfg": {
            "lr": 0.003,
            "clock_scale": 1.0
          },
          "mean": 0.7263866514
        },
        {
          "cfg": {
            "lr": 0.01,
            "clock_scale": 1.0
          },
          "mean": 0.7050461918
        }
      ]
    },
    "comparison": {
      "delta_mean": 0.529947238,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.543897301,
        0.5234872252,
        0.5733261853,
        0.4490493834,
        0.3902077377,
        0.5912492424,
        0.61420165,
        0.5541591793
      ],
      "p_value": 0.0081,
      "mde": 0.0629163503,
      "mde_rel_pct": 35.98432769,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "predicted_inactive_state_change": 0.0,
      "observed_baseline_inactive_change": 0.0,
      "observed_idea_inactive_change": 0.0,
      "observed_baseline_event_change": 3.5917565823,
      "observed_idea_event_change": 1.8746105433,
      "idea_active_fraction": 0.9478124976,
      "predicted_active_fraction_is_data_dependent": true,
      "confirmed": true
    },
    "protocol_notes": {
      "structural_match": "sequence-level temporal forecasting; recurrent state transition replacement",
      "baseline_method": "explicit residual Euler with one chronological transition per token",
      "idea_method": "implicit residual solve with event-clock deltas",
      "paired_seeds": [
        0,
        1,
        2,
        3,
        4,
        5,
        6,
        7
      ],
      "epochs": 15,
      "n_train": 400,
      "n_test": 200
    }
  },
  "how_to_run": "cd /home/maxwelhelp/all/math2nn/experiments/exp1400_stieltjes-event-driven-neural-state-laye && /home/maxwelhelp/main/bin/python3 bench_experiment.py",
  "files": [
    "bench_experiment.py",
    "artifacts/bench_report.json",
    "artifacts/run_output.json"
  ],
  "limitations": "Only the sequence track was tested; vision, tabular, and dynamics were not tested. The experiment used 400 training samples, 200 test samples, 15 epochs, a fixed threshold clock, and a damped fixed-point approximation rather than an exact nonlinear Newton solve. Wall-clock speed, memory use, custom reverse-adjoint implementation, and larger standard benchmark budgets were not evaluated.",
  "system_verdict": "partial",
  "practical_verdict": "harms",
  "mechanism_ok": 1,
  "system_judged": true
}