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

(универсальный)

Ответ агента:

{
  "worked": false,
  "confidence": 10,
  "verdict": "Built a Tangential Bellman Tie Resolver auxiliary training intervention on the structurally matched dynamics/rnn_small pendulum benchmark. The fixed-point iteration numerically contracted at the predicted beta bound (late ratio 0.729995 versus beta 0.73), but the idea scored substantially worse than the tuned baseline: MSE 0.788188 versus 0.0004558, with paired delta +0.787733 and permutation p=0.0081. Trained-model resolver weights had a very small mean branch span (0.00367), so the predicted continuous switching behavior was not confirmed; the system therefore did not work on this benchmark.",
  "metrics": {
    "baseline": "dynamics/rnn_small; 8-seed test MSE mean=0.0004557761, std=0.0000904587; best sweep cfg lr=0.005, weight_decay=0.0; sweep covered lr={0.001,0.003,0.005} and weight_decay={0,0.0001}",
    "idea": "dynamics/rnn_small; 8-seed test MSE mean=0.78818847, std=0.04192548; best idea cfg lr=0.003, weight_decay=0.0; paired delta (idea-baseline)=+0.78773270, p=0.0081, idea wins=0/8",
    "mechanism_signature": "Contraction ratio late=0.7299954 for beta=0.73; trained-model resolver weight span mean=0.0036721; hard-best frequency mean=0.4100; confirmed=false"
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_stage2.py",
  "files": [
    "bench_stage2.py",
    "bench_report.json",
    "bench_stdout.txt"
  ],
  "limitations": "Only the built-in dynamics track was tested, with 1,200 training and 400 test examples and 10 epochs rather than the harness defaults. The implementation is an auxiliary supervised counterfactual-branch loss rather than a complete continuous-action actor-critic or SAC integration with learned transition ensembles; no environment interaction, replay efficiency, FLOP, or larger-scale RL evaluation was tested.",
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.005,
        "wd": 0.0
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "wd": 0.0
          },
          "mean": 0.001154391880845651
        },
        {
          "cfg": {
            "lr": 0.001,
            "wd": 0.0001
          },
          "mean": 0.0011398583883419633
        },
        {
          "cfg": {
            "lr": 0.003,
            "wd": 0.0
          },
          "mean": 0.0006730457389494404
        },
        {
          "cfg": {
            "lr": 0.003,
            "wd": 0.0001
          },
          "mean": 0.00066975416848436
        },
        {
          "cfg": {
            "lr": 0.005,
            "wd": 0.0
          },
          "mean": 0.00047177909436868504
        },
        {
          "cfg": {
            "lr": 0.005,
            "wd": 0.0001
          },
          "mean": 0.0004883268411504105
        }
      ],
      "full": {
        "mean": 0.00045577605851576664,
        "std": 9.045873661427688e-05,
        "per_seed": [
          0.00045988193596713245,
          0.0005730087868869305,
          0.0003259392105974257,
          0.0005282864440232515,
          0.000366326974472031,
          0.00035140709951519966,
          0.0005543518927879632,
          0.0004870061238761991
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.7881884723901749,
      "std": 0.0419254847955854,
      "per_seed": [
        0.7438060641288757,
        0.8217167854309082,
        0.8174083828926086,
        0.857958197593689,
        0.8094730973243713,
        0.7287576198577881,
        0.7656309604644775,
        0.7607566714286804
      ],
      "n": 8
    },
    "comparison": {
      "delta_mean": 0.7877326963316591,
      "idea_wins": 0,
      "n_pairs": 8,
      "per_seed_diffs": [
        0.7433461821929086,
        0.8211437766440213,
        0.8170824436820112,
        0.8574299111496657,
        0.8091067703498993,
        0.7284062127582729,
        0.7650766085716896,
        0.7602696653048042
      ],
      "p_value": 0.0081,
      "mde": 0.03746007735504655,
      "mde_rel_pct": 8218.965576435758,
      "verdict": "idea worse (significant)",
      "system_worked": false
    },
    "mechanism_signature": {
      "math_check": {
        "operator_lipschitz_ratio": 0.14484610436506878,
        "beta": 0.73,
        "iteration_ratios": [
          0.4969973584994038,
          0.7098904435991482,
          0.7246338853808736,
          0.7286061045254636,
          0.7293810555254636,
          0.7299559137698793,
          0.729981689594991,
          0.7299953759512697
        ],
        "max_ratio_over_beta": 0.9999936656866709
      },
      "trained_behavior": {
        "resolver_weight_span": {
          "observed_mean": 0.0036720615407830337,
          "per_seed": [
            0.000908972229808569,
            0.006066256668418646,
            0.0012230807915329933,
            0.019085220992565155,
            0.0007767750066705048,
            0.0002523639996070415,
            0.000828575633931905,
            0.00023524700372945517
          ]
        },
        "hard_best_frequency": {
          "observed_mean": 0.4099999964237213,
          "per_seed": [
            0.0,
            1.0,
            0.7674999833106995,
            0.0,
            0.512499988079071,
            0.0,
            1.0,
            0.0
          ]
        },
        "prediction": "resolver branch weights vary continuously with counterfactual action consequences; measured on trained models",
        "confirmed": false
      },
      "idea_sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "wd": 0.0
          },
          "mean": 0.7889741063117981
        },
        {
          "cfg": {
            "lr": 0.003,
            "wd": 0.0
          },
          "mean": 0.7881884723901749
        },
        {
          "cfg": {
            "lr": 0.005,
            "wd": 0.0
          },
          "mean": 0.7896546497941017
        }
      ]
    },
    "protocol_notes": {
      "n_train": 1200,
      "n_test": 400,
      "epochs": 10,
      "structural_match": "controlled pendulum dynamics; action branches are final-step control perturbations",
      "baseline_knobs_swept": [
        "lr",
        "weight_decay"
      ]
    }
  },
  "system_verdict": "failed",
  "practical_verdict": "harms",
  "mechanism_ok": 0,
  "system_judged": true
}