{
  "id": 10211,
  "url": "https://arxiv.org/abs/2607.10805v1",
  "title": "Diagnosing and Mitigating Thinking Collapse in On-Policy Self-Distillation",
  "summary": "On-Policy Self-Distillation (OPSD) has emerged as a crucial paradigm for enhancing and aligning Large Language Models (LLMs). However, in complex reasoning tasks, OPSD paradoxically degrades downstream performance. In this paper, we systematically investigate this pathology and identify a severe optimization trap we define as \\textbf{Thinking Collapse} -- a sharp decline in the model's native intermediate reasoning behavior, measured by epistemic-token density (ET per 1k). Through entropy-based ",
  "authors": "Keqin Peng, Chen Li, Yuanxin Ouyang, Yancheng Yuan, Liang Ding",
  "category": "research",
  "topics": "regulation,healthcare,finance-investment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-12T15:24:25.000Z",
  "fetched_at": "2026-07-14T16:55:59.928Z",
  "source_slug": "arxiv-cslg",
  "source_name": "arXiv cs.LG",
  "source_homepage": "https://arxiv.org/list/cs.LG/recent",
  "ethics_ai_record_url": "https://ethics.ai/record/10211",
  "original_url": "https://arxiv.org/abs/2607.10805v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}