{
  "bench_version": 1,
  "track": "custom_stochastic_expectation",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.006
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.0015
        },
        "mean": 0.011956734815612435
      },
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 0.013272208627313375
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 0.010816274618264288
      }
    ],
    "full": {
      "mean": 0.009381077397847548,
      "std": 0.005076745501359762,
      "per_seed": [
        0.021740078926086426,
        0.0081116808578372,
        0.009601366706192493,
        0.0038119719829410315,
        0.006551957223564386,
        0.006829647812992334,
        0.010959993116557598,
        0.007441922556608915
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 0.00570399034768343,
    "std": 0.002055810761713156,
    "per_seed": [
      0.006588147487491369,
      0.010348089039325714,
      0.0069997236132621765,
      0.00420125387609005,
      0.003918817266821861,
      0.00435355631634593,
      0.004861464258283377,
      0.00436087092384696
    ],
    "n": 8
  },
  "comparison": {
    "delta_mean": -0.0036770870501641184,
    "idea_wins": 6,
    "n_pairs": 8,
    "per_seed_diffs": [
      -0.015151931438595057,
      0.002236408181488514,
      -0.002601643092930317,
      0.0003892818931490183,
      -0.002633139956742525,
      -0.0024760914966464043,
      -0.006098528858274221,
      -0.0030810516327619553
    ],
    "p_value": 0.04775,
    "mde": 0.004391476661630811,
    "mde_rel_pct": 46.81207152856897,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "mechanism_signature": {
    "track_structure": "explicit smooth Gaussian expectation in each minibatch",
    "gradient_variance": {
      "n": 32,
      "repetitions": 32,
      "predicted": "lower gradient variance for smooth base-noise integrand",
      "gradient_variance_iid": 5.355114990379661e-05,
      "gradient_variance_qmc": 1.6305946701322682e-05,
      "variance_reduction": 0.695507066970253,
      "confirmed": true
    }
  },
  "custom_track": {
    "name": "stochastic_expectation",
    "file": "qmc_bench.py",
    "domain": "stochastic_expectation"
  },
  "idea_sweep": [
    {
      "cfg": {
        "lr": 0.0015
      },
      "result": {
        "mean": 0.0069728404632769525,
        "std": 0.0035963506180249297,
        "per_seed": [
          0.005615524481981993,
          0.013930782675743103,
          0.01141586434096098,
          0.00786577071994543,
          0.005591781344264746,
          0.00341523508541286,
          0.004224354401230812,
          0.0037234106566756964
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.003
      },
      "result": {
        "mean": 0.00570399034768343,
        "std": 0.002055810761713156,
        "per_seed": [
          0.006588147487491369,
          0.010348089039325714,
          0.0069997236132621765,
          0.00420125387609005,
          0.003918817266821861,
          0.00435355631634593,
          0.004861464258283377,
          0.00436087092384696
        ],
        "n": 8
      }
    },
    {
      "cfg": {
        "lr": 0.006
      },
      "result": {
        "mean": 0.00680300744716078,
        "std": 0.004497406606670575,
        "per_seed": [
          0.010847600176930428,
          0.003560971235856414,
          0.017018528655171394,
          0.004085659049451351,
          0.005837064236402512,
          0.004683156963437796,
          0.0027501408476382494,
          0.0056409384123981
        ],
        "n": 8
      }
    }
  ],
  "protocol": {
    "paired_seeds": [
      0,
      1,
      2,
      3,
      4,
      5,
      6,
      7
    ],
    "epochs": 18,
    "batch": 32,
    "lr_union_both_sides": [
      0.0015,
      0.003,
      0.006
    ],
    "model_parity": true
  }
}