{
  "id": "2602.16490",
  "title": "From Growing to Looping: A Unified View of Iterative Computation in LLMs",
  "authors": "Ferdinand Kapl, Emmanouil Angelis, Kaitlin Maile, Johannes von Oswald, Stefan Bauer",
  "first_public": "2026-02-18",
  "added_at": "2026-11-03",
  "venue": "arXiv",
  "arxiv": "2602.16490",
  "paper_url": "https://arxiv.org/abs/2602.16490",
  "html_url": "https://arxiv.org/html/2602.16490v1",
  "lens": [
    "recurrent",
    "analysis"
  ],
  "representation": [
    "hidden-state"
  ],
  "computation": [
    "recurrent",
    "depth-wise"
  ],
  "training": [
    "pretraining",
    "supervised",
    "analysis-only"
  ],
  "questions": [
    "architecture",
    "interpretability",
    "efficiency"
  ],
  "evidence": [
    "empirical",
    "mechanistic"
  ],
  "modality": [
    "language",
    "symbolic"
  ],
  "summary": "Compares depth growing and weight-tied loops using matched model budgets and interventions on intermediate layers. Finds related patterns of repeated computation and improves reasoning by adding middle-block loops to depth-grown models during inference or continued training.",
  "provenance": "Primary-source literature review, 2026-11-03",
  "author_names": [
    "Ferdinand Kapl",
    "Emmanouil Angelis",
    "Kaitlin Maile",
    "Johannes von Oswald",
    "Stefan Bauer"
  ],
  "record_url": "https://latent-implicit-thinking.github.io/papers/2602.16490/",
  "legacy_record_url": "https://latent-implicit-thinking.github.io/papers/#paper-2602-16490",
  "citation_date": "2026-02-18",
  "json_url": "https://latent-implicit-thinking.github.io/papers/2602.16490/record.json",
  "markdown_url": "https://latent-implicit-thinking.github.io/papers/2602.16490/index.md",
  "bibtex_url": "https://latent-implicit-thinking.github.io/papers/2602.16490/citation.bib",
  "csl_url": "https://latent-implicit-thinking.github.io/papers/2602.16490/citation.csl.json",
  "pdf_url": "https://arxiv.org/pdf/2602.16490"
}