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

(универсальный)

Ответ агента:

{
  "worked": true,
  "confidence": 8,
  "verdict": "The reversible 3-state Mealy mixer was evaluated on the structurally matched controlled-pendulum dynamics track against the standard rnn_small GRU under paired seeds and equal training budgets. It significantly improved test MSE: 0.0006506 versus 0.0014198, delta_mean=-0.0007692 with p=0.0163 and 7/8 wins. The trained-model conservation signature was also confirmed exactly, although memory savings and long-sequence scaling were not tested.",
  "metrics": {
    "baseline": "Mean test MSE 0.0014198302 ± 0.0003983911; tuned lr=0.01, epochs=12.",
    "idea": "Mean test MSE 0.0006506168 ± 0.0003999347; lr=0.01, scale=0.5, epochs=12; paired delta=-0.0007692134, p=0.0163."
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_experiment.py",
  "files": [
    "bench_experiment.py",
    "bench_report.json"
  ],
  "limitations": "Only the built-in dynamics track was tested using 400 training and 200 test examples with an 8-step window. Activation memory, reversible backpropagation, sequence lengths 512-8192, language modeling, GPU throughput, and broader architecture validation were not measured. The implementation uses hard binary thresholding and is an MVP.",
  "bench_report": {
    "bench_version": 1,
    "track": "dynamics",
    "model": "rnn_small",
    "metric_direction": "lower is better",
    "n_seeds": 8,
    "baseline": {
      "best_cfg": {
        "lr": 0.01,
        "epochs": 12
      },
      "sweep": [
        {
          "cfg": {
            "lr": 0.001,
            "epochs": 12
          },
          "mean": 0.004298779065720737
        },
        {
          "cfg": {
            "lr": 0.003,
            "epochs": 12
          },
          "mean": 0.0034615940821822733
        },
        {
          "cfg": {
            "lr": 0.01,
            "epochs": 12
          },
          "mean": 0.0014865802077110857
        }
      ],
      "full": {
        "mean": 0.0014198302378645167,
        "std": 0.00039839114901047844,
        "per_seed": [
          0.0009327387670055032,
          0.0018641706556081772,
          0.0010785290505737066,
          0.0020708823576569557,
          0.0009260809747502208,
          0.0014055331703275442,
          0.0016443432541564107,
          0.001436363672837615
        ],
        "n": 8
      }
    },
    "idea": {
      "mean": 0.0006506168301712023,
      "std": 0.00039993469646715593,
      "per_seed": [
        0.00010142059181816876,
        0.0005146179464645684,
        0.001238752854987979,
        0.001161463325843215,
        0.0002165312325814739,
        0.0007791328243911266,
        0.0008627780480310321,
        0.0003302378172520548
      ],
      "n": 8,
      "selected_cfg": {
        "lr": 0.01,
        "epochs": 12,
        "scale": 0.5
      }
    },
    "comparison": {
      "delta_mean": -0.0007692134076933144,
      "idea_wins": 7,
      "n_pairs": 8,
      "per_seed_diffs": [
        -0.0008313181751873344,
        -0.0013495527091436088,
        0.0001602238044142723,
        -0.0009094190318137407,
        -0.0007095497421687469,
        -0.0006264003459364176,
        -0.0007815652061253786,
        -0.0011061258555855602
      ],
      "p_value": 0.0163,
      "mde": 0.000368799706207264,
      "mde_rel_pct": 25.974915618218837,
      "verdict": "idea better (significant)",
      "system_worked": true
    },
    "math_check": {
      "local_bijective": true,
      "local_conservation_errors": [
        0,
        0,
        0,
        0,
        0,
        0
      ],
      "max_abs_conservation_error": 0
    },
    "idea_sweep": [
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 12,
          "scale": 0.5
        },
        "result": {
          "mean": 0.0006506168301712023,
          "std": 0.00039993469646715593,
          "per_seed": [
            0.00010142059181816876,
            0.0005146179464645684,
            0.001238752854987979,
            0.001161463325843215,
            0.0002165312325814739,
            0.0007791328243911266,
            0.0008627780480310321,
            0.0003302378172520548
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 12,
          "scale": 1.0
        },
        "result": {
          "mean": 0.0011277017038082704,
          "std": 0.0005102781821450115,
          "per_seed": [
            0.0007301242440007627,
            0.001474376767873764,
            0.0014355103485286236,
            0.0018912762170657516,
            0.0006100746686570346,
            0.0010276326211169362,
            0.0015459696296602488,
            0.0003066491335630417
          ],
          "n": 8
        }
      },
      {
        "cfg": {
          "lr": 0.01,
          "epochs": 12,
          "scale": 2.0
        },
        "result": {
          "mean": 0.0017072468253900297,
          "std": 0.0008234145200694133,
          "per_seed": [
            0.0011916327057406306,
            0.002574339509010315,
            0.0014427917776629329,
            0.0016165744746103883,
            0.0012496613198891282,
            0.002416426781564951,
            0.0029221479780972004,
            0.0002444000565446913
          ],
          "n": 8
        }
      }
    ],
    "structural_match": "controlled damped pendulum rollout tests stability/control of sequential state evolution",
    "mechanism_signature": {
      "quantity": "symbol_weight + carrier_weight conservation over trained model tokens",
      "predicted": 0.0,
      "observed": 0.0,
      "absolute_error": 0.0,
      "confirmed": true
    }
  },
  "system_verdict": "worked",
  "practical_verdict": "helps",
  "mechanism_ok": 1,
  "system_judged": true
}