logbook.json

{
  "schema_version": 1,
  "title": "Repro — CapBencher: Give Your LLM Benchmark a Built-in Alarm for Test-Set Overfitting (formerly “How Can I Publish My LLM Benchmark Without Giving the True Answers Away?”)",
  "emoji": "🎯",
  "space_id": "Boopster/oCNT5PcMSQ",
  "paper": {
    "arxiv_id": "2505.18102"
  },
  "tags": [
    "icml2026-repro",
    "paper-oCNT5PcMSQ"
  ],
  "updated_at": "2026-07-17T23:21:31+00:00",
  "root": {
    "slug": "index",
    "title": "Repro — CapBencher: Give Your LLM Benchmark a Built-in Alarm for Test-Set Overfitting (formerly “How Can I Publish My LLM Benchmark Without Giving the True Answers Away?”)",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "claim-a-detecting-deliberate-contamination",
        "title": "Official claim 1 — detecting deliberate contamination",
        "file": "pages/claim-a-detecting-deliberate-contamination/page.md",
        "children": []
      },
      {
        "slug": "claim-b-the-bayes-accuracy-cap",
        "title": "Official claim 2 — the Bayes-accuracy cap",
        "file": "pages/claim-b-the-bayes-accuracy-cap/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "Conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      },
      {
        "slug": "further-evidence-ranking-and-merge",
        "title": "Further evidence — BF16 ranking ladder and merge detection",
        "file": "pages/further-evidence-ranking-and-merge/page.md",
        "children": []
      },
      {
        "slug": "methods-and-provenance",
        "title": "Methods and provenance",
        "file": "pages/methods-and-provenance/page.md",
        "children": []
      },
      {
        "slug": "run-history-and-compute",
        "title": "Run history and compute",
        "file": "pages/run-history-and-compute/page.md",
        "children": []
      }
    ]
  },
  "agent_view_tokens": 19484,
  "revision": "1784330491786477000"
}