{
  "bench_version": 1,
  "track": "dynamics",
  "model": "rnn_small",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.01,
      "M": 1
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "M": 1
        },
        "mean": 0.001921573159052059
      },
      {
        "cfg": {
          "lr": 0.003,
          "M": 1
        },
        "mean": 0.0011179699067724869
      },
      {
        "cfg": {
          "lr": 0.01,
          "M": 1
        },
        "mean": 0.0006299250526353717
      }
    ],
    "full": {
      "mean": 0.0005546553857129766,
      "std": 0.00021396768993230646,
      "per_seed": [
        0.0004971042508259416,
        0.0004999673110432923,
        0.00046853424282744527,
        0.0010540944058448076,
        0.0005468997405841947,
        0.00023900401720311493,
        0.0006055250996723771,
        0.0005261140177026391
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.0005546553857129766,
    "std": 0.00021396768993230646,
    "per_seed": [
      0.0004971042508259416,
      0.0004999673110432923,
      0.00046853424282744527,
      0.0010540944058448076,
      0.0005468997405841947,
      0.00023900401720311493,
      0.0006055250996723771,
      0.0005261140177026391
    ],
    "n": 8,
    "sweep": [
      {
        "cfg": {
          "lr": 0.001,
          "M": 1
        },
        "mean": 0.001921573159052059
      },
      {
        "cfg": {
          "lr": 0.003,
          "M": 1
        },
        "mean": 0.0011179699067724869
      },
      {
        "cfg": {
          "lr": 0.01,
          "M": 1
        },
        "mean": 0.0006299250526353717
      }
    ]
  },
  "comparison": {
    "delta_mean": 0.0,
    "idea_wins": 0,
    "n_pairs": 8,
    "per_seed_diffs": [
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0,
      0.0
    ],
    "p_value": 1.0,
    "mde": 0.0,
    "mde_rel_pct": 0.0,
    "verdict": "no measurable effect",
    "system_worked": false
  },
  "mechanism_signature": {
    "prediction": "Lyapunov certificate filters sampled updates; rejected proposals should have positive certificate and accepted proposals nonpositive.",
    "trained_model_measurements": [
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.49112362704343265
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.4969692817992634
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.5120618219176928
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.5117074669235282
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.5226765928996934
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.48182762745353913
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.4946398569477929
      },
      {
        "accepted": 72,
        "rejected": 0,
        "checks": 72,
        "violations": 0,
        "violation_rate": 0.0,
        "mean_certificate": -0.508300371054146
      }
    ],
    "confirmed": false
  },
  "protocol_notes": "Baseline and idea share rnn_small, data, epochs, batch, Adam, and lr/M grid; only certificate rejection differs."
}