{
  "abstract": "Universal machine-learning interatomic potentials (u-MLIPs) aim to generalize across diverse configurations. Benchmarks enable reproducible evaluation but may not expose failures outside their predefined scope. Here, we show that physics-informed search can complement benchmark-based evaluation by uncovering hidden failure modes. We introduce MLIP Detective, an agentic framework for active failure mode discovery. Starting from benchmark evidence, MLIP Detective generates falsifiable, physics-informed failure hypotheses, screens them with inexpensive simulations, and escalates only the most suspicious cases to human experts together with proposed verification protocols. Without issue-specific prompting, MLIP Detective identified and characterized a systematic anomaly in MACE-MPA-0: the model predicted some relaxed adsorbate-surface systems involving O- or F-containing adsorbates to be higher in energy than their corresponding separated fragments. Using cross-model comparisons, MLIP Detective further inferred a likely training-data origin for the anomaly, consistent with recent reports.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2609.08399",
  "author_total": 4,
  "authors": [
    {
      "name": "Ryuhei Okuno"
    },
    {
      "name": "Nontawat Charoenphakdee"
    },
    {
      "name": "Kaoru Hisama"
    },
    {
      "name": "Yuta Tsuboi"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/35b5nt0ta69v2.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2609.08399",
  "fulltext_url": null,
  "id": "35b5nt0ta69v2",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2609.08399",
  "language": null,
  "licence": "CC-BY-4.0",
  "licence_label": "CC BY 4.0",
  "licence_source": "article",
  "licence_url": "http://creativecommons.org/licenses/by/4.0/",
  "metadata_licence": "CC0-1.0",
  "native_subjects": [
    [
      "arxiv",
      "cond-mat.mtrl-sci"
    ],
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "physics.chem-ph"
    ],
    [
      "openalex",
      "2505|Machine Learning in Materials Science"
    ]
  ],
  "openalex": "W7212074203",
  "pdf_url": "https://arxiv.org/pdf/2609.08399",
  "pmcid": null,
  "pmid": null,
  "published": "2026-09-08",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/35b5nt0ta69v2.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:19:49Z",
      "native_id": "2609.08399",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2609.08399"
    }
  ],
  "subjects": [
    {
      "field": "chemistry",
      "origin": "arxiv:physics.chem-ph",
      "subfield": "physical"
    },
    {
      "field": "physics",
      "origin": "arxiv:cond-mat.mtrl-sci",
      "subfield": "condensed-matter"
    }
  ],
  "title": "MLIP Detective: Active Failure Mode Discovery Beyond Benchmark Scores for Machine-Learning Interatomic Potentials",
  "title_text": "MLIP Detective: Active Failure Mode Discovery Beyond Benchmark Scores for Machine-Learning Interatomic Potentials",
  "type": "preprint",
  "updated_at": "2026-10-30T12:19:49Z",
  "url": "https://research.draconitic.com/paper/35b5nt0ta69v2",
  "venue": "arXiv",
  "versions": []
}