{
  "id": "2611.07211",
  "title": "Not All Loops Cost the Same: Phase-Dependent Depth for Serving Looped Transformers",
  "authors": "Zhixu Du, Yuhan Cheng, Hai \"Helen\" Li, Yiran Chen",
  "first_public": "2026-11-04",
  "added_at": "2026-11-09",
  "venue": "arXiv",
  "arxiv": "2611.07211",
  "paper_url": "https://arxiv.org/abs/2611.07211",
  "html_url": "https://arxiv.org/html/2611.07211v1",
  "lens": [
    "recurrent"
  ],
  "representation": [
    "hidden-state",
    "cache-memory"
  ],
  "computation": [
    "recurrent"
  ],
  "training": [
    "inference-only"
  ],
  "questions": [
    "efficiency",
    "controllability",
    "evaluation"
  ],
  "evidence": [
    "empirical"
  ],
  "modality": [
    "language"
  ],
  "summary": "Uses fewer recurrent steps for prompt tokens while retaining more computation for prompt endpoints and generated tokens. Removes finished positions from later loop calls and adjusts prompt depth as the serving queue grows.",
  "provenance": "Primary-source literature review, 2026-11-09",
  "author_names": [
    "Zhixu Du",
    "Yuhan Cheng",
    "Hai \"Helen\" Li",
    "Yiran Chen"
  ],
  "record_url": "https://latent-implicit-thinking.github.io/papers/2611.07211/",
  "legacy_record_url": "https://latent-implicit-thinking.github.io/papers/#paper-2611-07211",
  "citation_date": "2026-11-04",
  "json_url": "https://latent-implicit-thinking.github.io/papers/2611.07211/record.json",
  "markdown_url": "https://latent-implicit-thinking.github.io/papers/2611.07211/index.md",
  "bibtex_url": "https://latent-implicit-thinking.github.io/papers/2611.07211/citation.bib",
  "csl_url": "https://latent-implicit-thinking.github.io/papers/2611.07211/citation.csl.json",
  "pdf_url": "https://arxiv.org/pdf/2611.07211"
}