{
  "bench_version": 1,
  "track": "tabular",
  "model": "mlp_tiny",
  "metric_direction": "lower is better",
  "n_seeds": 8,
  "baseline": {
    "best_cfg": {
      "lr": 0.009
    },
    "sweep": [
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 225.13445281982422
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 214.2943878173828
      },
      {
        "cfg": {
          "lr": 0.009
        },
        "mean": 201.62980270385742
      }
    ],
    "full": {
      "mean": 200.0606346130371,
      "std": 11.077029829348923,
      "per_seed": [
        212.36439514160156,
        200.49159240722656,
        205.28561401367188,
        188.3776092529297,
        203.29046630859375,
        185.93289184570312,
        187.44577026367188,
        217.29673767089844
      ],
      "n": 8
    }
  },
  "idea": {
    "mean": 185.36586952209473,
    "std": 11.095241493609498,
    "per_seed": [
      195.4981231689453,
      187.09390258789062,
      192.18927001953125,
      175.18148803710938,
      185.1798553466797,
      171.63917541503906,
      171.85842895507812,
      204.28671264648438
    ],
    "n": 8,
    "cfg": {
      "lr": 0.009
    },
    "runs": [
      {
        "metric": 195.4981231689453,
        "switched": true,
        "switch_step": 23,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.01804690866414924,
        "predicted_minus_q2": -382299857278.8767,
        "q_tail_mean": 604388.9543802879,
        "device": "cuda"
      },
      {
        "metric": 187.09390258789062,
        "switched": true,
        "switch_step": 30,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.009911357660239538,
        "predicted_minus_q2": -298791806484.8164,
        "q_tail_mean": 499215.2766124988,
        "device": "cuda"
      },
      {
        "metric": 192.18927001953125,
        "switched": true,
        "switch_step": 30,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.06313450122089037,
        "predicted_minus_q2": -307170570800.53937,
        "q_tail_mean": 521645.67662283836,
        "device": "cuda"
      },
      {
        "metric": 175.18148803710938,
        "switched": true,
        "switch_step": 24,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": 0.0313864532309592,
        "predicted_minus_q2": -277706767948.9584,
        "q_tail_mean": 508872.9958189457,
        "device": "cuda"
      },
      {
        "metric": 185.1798553466797,
        "switched": true,
        "switch_step": 25,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.08906700192000473,
        "predicted_minus_q2": -469862343448.6883,
        "q_tail_mean": 618141.8733281547,
        "device": "cuda"
      },
      {
        "metric": 171.63917541503906,
        "switched": true,
        "switch_step": 23,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.07050379253268516,
        "predicted_minus_q2": -322708159123.9784,
        "q_tail_mean": 542024.741677214,
        "device": "cuda"
      },
      {
        "metric": 171.85842895507812,
        "switched": true,
        "switch_step": 28,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.05052109490887936,
        "predicted_minus_q2": -508205333696.0094,
        "q_tail_mean": 697474.1491504112,
        "device": "cuda"
      },
      {
        "metric": 204.28671264648438,
        "switched": true,
        "switch_step": 28,
        "lr_final": 0.013499999999999998,
        "clip_steps": 72,
        "observed_log_gap_slope": -0.04710672958307441,
        "predicted_minus_q2": -307426876057.59827,
        "q_tail_mean": 536556.6369037025,
        "device": "cuda"
      }
    ],
    "sweep": [
      {
        "cfg": {
          "lr": 0.003
        },
        "mean": 222.8292999267578
      },
      {
        "cfg": {
          "lr": 0.006
        },
        "mean": 206.63719177246094
      },
      {
        "cfg": {
          "lr": 0.009
        },
        "mean": 187.49069595336914
      }
    ]
  },
  "comparison": {
    "delta_mean": -14.694765090942383,
    "idea_wins": 8,
    "n_pairs": 8,
    "per_seed_diffs": [
      -16.86627197265625,
      -13.397689819335938,
      -13.096343994140625,
      -13.196121215820312,
      -18.110610961914062,
      -14.293716430664062,
      -15.58734130859375,
      -13.010025024414062
    ],
    "p_value": 0.0081,
    "mde": 1.6318810404189403,
    "mde_rel_pct": 0.8156932239945007,
    "verdict": "idea better (significant)",
    "system_worked": true
  },
  "mechanism_signature": {
    "quantity": "tail log(training loss gap) slope versus -mean(q^2), measured during trained NN runs",
    "observed_mean": -0.039613116657370454,
    "predicted_mean": -359271464354.93317,
    "relative_error_mean": 0.9999999999998959,
    "tolerance": 0.2,
    "confirmed": false,
    "switch_fraction": 1.0
  },
  "audit": {
    "baseline_lr_grid": [
      0.003,
      0.006,
      0.009
    ],
    "idea_lr_grid": [
      0.003,
      0.006,
      0.009
    ],
    "baseline_selected_lr": 0.009,
    "idea_selected_lr": 0.009,
    "epochs": 18,
    "batch": 128,
    "clip_norm": 1.0,
    "domain_rationale": "tabular is the prescribed structural track for optimizer and learning-rate schedule ideas"
  }
}