{
  "abstract": "When a language model is distilled from a teacher that shares its base parameters, the student can inherit behavioural traits of the teacher---including misalignment---even from data that is semantically unrelated to the trait (Cloud et al., 2026). All existing defences require knowing the trait. We remove this requirement and prove that a subliminally transferred trait can be recovered from the parameter updates alone, without supervision, by exploiting a single asymmetry: the trait is expressed consistently across diverse datasets generated by the teacher, whereas task-specific and shared capabilities vary. The resulting estimator returns a weight-space direction that can be amplified to discover and interpret the trait or projected out to suppress it. Applied to the subliminal-learning benchmarks of (Cloud et al., 2026), our method recovers the transmitted animal-preference traits without any trait labels and suppresses their expression with almost no impact to task performance.",
  "abstract_shown": true,
  "affiliations": [],
  "article_type": null,
  "arxiv": "2610.13536",
  "author_total": 4,
  "authors": [
    {
      "name": "Joseph Egan"
    },
    {
      "name": "Ciarán M. Gilligan-Lee"
    },
    {
      "name": "Michael O'Riordan"
    },
    {
      "name": "Yuchen Zhu"
    }
  ],
  "bibtex": "https://research.draconitic.com/paper/5x21maqwwdegy.bib",
  "cited_by": 0,
  "doi": "10.48550/arxiv.2610.13536",
  "fulltext_url": null,
  "id": "5x21maqwwdegy",
  "issns": [],
  "landing_url": "https://arxiv.org/abs/2610.13536",
  "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",
      "cs.AI"
    ],
    [
      "arxiv",
      "cs.LG"
    ],
    [
      "arxiv",
      "stat.ML"
    ],
    [
      "openalex",
      "1702|Adversarial Robustness in Machine Learning"
    ]
  ],
  "openalex": "W7221510522",
  "pdf_url": "https://arxiv.org/pdf/2610.13536",
  "pmcid": null,
  "pmid": null,
  "published": "2026-10-09",
  "publisher": null,
  "ris": "https://research.draconitic.com/paper/5x21maqwwdegy.ris",
  "sources": [
    {
      "first_seen": "2026-10-30T12:28:46Z",
      "native_id": "2610.13536",
      "source": "arxiv",
      "url": "https://arxiv.org/abs/2610.13536"
    }
  ],
  "subjects": [
    {
      "field": "mathematics",
      "origin": "arxiv:stat.ML",
      "subfield": "statistics"
    }
  ],
  "title": "Defence against the dark arts of distillation: Unsupervised discovery of subliminal learning",
  "title_text": "Defence against the dark arts of distillation: Unsupervised discovery of subliminal learning",
  "type": "preprint",
  "updated_at": "2026-10-30T12:28:46Z",
  "url": "https://research.draconitic.com/paper/5x21maqwwdegy",
  "venue": "arXiv",
  "versions": []
}