{
  "frozen_at": "2026-07-18T23:46:57.951556+00:00",
  "git_sha": "f93f4b5578c221d3d2665f9b305f3e86759be862+dirty",
  "spec": {
    "analysis": "campaign.ppe.analyze",
    "hypotheses": [
      {
        "id": "H-tail-plateau",
        "prediction": {
          "ci_excludes": 0.0,
          "comparator": ">",
          "effect": null,
          "metric": "spearman_tail_vs_bon_plateau",
          "threshold": 0.3
        },
        "scoreboard_row": null,
        "statement": "the tail index predicts where best-of-k stops helping on the five PPE verifiable sets"
      },
      {
        "id": "H-lowerquantile",
        "prediction": {
          "ci_excludes": null,
          "comparator": "==",
          "effect": null,
          "metric": "bon_lowerquantile_beats_mean",
          "threshold": 1.0
        },
        "scoreboard_row": null,
        "statement": "the lower-quantile best-of-n aggregation beats the mean against human labels, reproducing the PPE finding"
      }
    ],
    "id": "campaign-ppe-bon",
    "kill_criteria": [
      {
        "comparator": "<=",
        "id": "K-ppe",
        "metric": "spearman_tail_vs_bon_plateau_ci_low",
        "threshold": 0.0
      }
    ],
    "science": "S03-thermo",
    "subjects": {
      "datasets": [
        "ppe-best-of-k",
        "ppe-human"
      ],
      "extra": {
        "slice_hashes": {
          "ppe-best-of-k": "ds:43a17b70ed1c3ee83423bb4df2cce4ac",
          "ppe-human": "ds:f9fa29ebabe6994fded4e9057ec07c10"
        }
      },
      "organisms": [],
      "signals": [
        "skywork-v2-qwen3-8b",
        "skywork-v2-llama31-8b",
        "skywork-v2-qwen3-0.6b"
      ]
    },
    "title": "Tail index forecasts the best-of-k plateau on PPE",
    "version": 1
  },
  "spec_hash": "spec:7077cf849ed8492fc327d970aa2ede55",
  "study_id": "study:campaign-ppe-bon@v1#7077cf84"
}