{
  "id": "2610.31421",
  "title": "How Depth Growth and Looping Shape In-Context Learning in Language Models",
  "authors": "Ferdinand Kapl, Kaitlin Maile, Stefan Bauer, Johannes von Oswald",
  "first_public": "2026-10-22",
  "added_at": "2026-11-02",
  "venue": "arXiv",
  "arxiv": "2610.31421",
  "paper_url": "https://arxiv.org/abs/2610.31421",
  "html_url": "https://arxiv.org/html/2610.31421v1",
  "lens": [
    "recurrent",
    "analysis"
  ],
  "representation": [
    "hidden-state"
  ],
  "computation": [
    "depth-wise",
    "recurrent"
  ],
  "training": [
    "pretraining",
    "analysis-only"
  ],
  "questions": [
    "architecture",
    "capability",
    "interpretability"
  ],
  "evidence": [
    "empirical",
    "mechanistic"
  ],
  "modality": [
    "language",
    "symbolic",
    "vision"
  ],
  "summary": "Compares dense, progressively grown and looped Transformers on in-context learning. Attention-head tracing and ablation link stronger context use in grown and looped models to greater reliance on induction heads.",
  "provenance": "Primary-source literature review, 2026-11-02",
  "author_names": [
    "Ferdinand Kapl",
    "Kaitlin Maile",
    "Stefan Bauer",
    "Johannes von Oswald"
  ],
  "record_url": "https://latent-implicit-thinking.github.io/papers/2610.31421/",
  "legacy_record_url": "https://latent-implicit-thinking.github.io/papers/#paper-2610-31421",
  "citation_date": "2026-10-22",
  "json_url": "https://latent-implicit-thinking.github.io/papers/2610.31421/record.json",
  "markdown_url": "https://latent-implicit-thinking.github.io/papers/2610.31421/index.md",
  "bibtex_url": "https://latent-implicit-thinking.github.io/papers/2610.31421/citation.bib",
  "csl_url": "https://latent-implicit-thinking.github.io/papers/2610.31421/citation.csl.json",
  "pdf_url": "https://arxiv.org/pdf/2610.31421"
}