{
  "id": "2611.04663",
  "title": "Best-of-Depth: Learning Exploratory Dynamics in Looped Transformers",
  "authors": "Shota Takashiro, Takeshi Kojima, Masanori Koyama, Kohei Hayashi, Yusuke Iwasawa",
  "first_public": "2026-11-03",
  "added_at": "2026-11-09",
  "venue": "arXiv",
  "arxiv": "2611.04663",
  "paper_url": "https://arxiv.org/abs/2611.04663",
  "html_url": "https://arxiv.org/html/2611.04663v1",
  "lens": [
    "recurrent"
  ],
  "representation": [
    "hidden-state"
  ],
  "computation": [
    "recurrent"
  ],
  "training": [
    "supervised"
  ],
  "questions": [
    "capability",
    "scaling",
    "theory"
  ],
  "evidence": [
    "empirical",
    "formal"
  ],
  "modality": [
    "language",
    "symbolic"
  ],
  "summary": "Trains a looped transformer using the lowest-loss prediction along its recurrent trajectory. Different depths can produce different valid solutions, which are evaluated on puzzles and code generation.",
  "provenance": "Primary-source literature review, 2026-11-09",
  "author_names": [
    "Shota Takashiro",
    "Takeshi Kojima",
    "Masanori Koyama",
    "Kohei Hayashi",
    "Yusuke Iwasawa"
  ],
  "record_url": "https://latent-implicit-thinking.github.io/papers/2611.04663/",
  "legacy_record_url": "https://latent-implicit-thinking.github.io/papers/#paper-2611-04663",
  "citation_date": "2026-11-03",
  "json_url": "https://latent-implicit-thinking.github.io/papers/2611.04663/record.json",
  "markdown_url": "https://latent-implicit-thinking.github.io/papers/2611.04663/index.md",
  "bibtex_url": "https://latent-implicit-thinking.github.io/papers/2611.04663/citation.bib",
  "csl_url": "https://latent-implicit-thinking.github.io/papers/2611.04663/citation.csl.json",
  "pdf_url": "https://arxiv.org/pdf/2611.04663"
}