{
  "abstract": "We study whether benchmarks reliably guide model selection as developers adapt to evaluation feedback across multiple criteria. We find that the worst-case test-set size needed to estimate the best score among $k$ adaptively chosen models, under any convex combination of the criteria, grows exponentially with the number of criteria, reaching the $Θ(\\sqrt{k})$ cost of answering $k$ adaptive statistical queries with only $O(\\log k)$ criteria, at fixed accuracy and confidence. In attacks on multi-task large language model benchmarks with five to ten criteria, feedback restricted to nondominated task profiles produces large reused-to-held-out score gaps and frequent false winners. These results challenge a prominent explanation for prior observed reliable benchmark reuse---that developers mainly respond to convincing improvements over the current best---in rich-feedback settings, while leaving open how often ordinary model development encounters this vulnerability.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.32109",
  "author_total": 2,
  "authors": [
    {
      "name": "Youssef Allouah"
    },
    {
      "name": "John Duchi"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/b1t2fcz8jcetw.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.32109",
  "fulltext_url": null,
  "id": "b1t2fcz8jcetw",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2609.32109",
  "language": null,
  "licence": "LicenseRef-arXiv-nonexclusive-1.0",
  "licence_label": "arXiv licence",
  "licence_source": "article",
  "licence_url": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "openalex",
      "1702|Machine Learning and Data Classification"
    ]
  ],
  "openalex": "W7214735252",
  "pdf_url": "https://arxiv.org/pdf/2609.32109",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-26",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/b1t2fcz8jcetw.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:25:26Z",
      "native_id": "2609.32109",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.32109"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ML",
      "subfield": "statistics"
    }
  ],
  "title": "How Reusable Are Benchmarks with Richer Feedback?",
  "title_text": "How Reusable Are Benchmarks with Richer Feedback?",
  "type": "preprint",
  "updated_at": "2026-10-30T12:25:26Z",
  "url": "https://research.draconitic.com/paper/b1t2fcz8jcetw",
  "venue": "arXiv",
  "versions": []
}