{"ID":2869174,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.14662","arxiv_id":"2509.14662","title":"Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld's Episode Theory","abstract":"While Large Reasoning Models (LRMs) generate extensive chain-of-thought reasoning, we lack a principled framework for understanding how these thoughts are structured. In this paper, we introduce a novel approach by applying Schoenfeld's Episode Theory, a classic cognitive framework for human mathematical problem-solving, to analyze the reasoning traces of LRMs. We annotated thousands of sentences and paragraphs from model-generated solutions to math problems using seven cognitive labels (e.g., Plan, Implement, Verify). The result is the first publicly available benchmark for the fine-grained analysis of machine reasoning, including a large annotated corpus and detailed annotation guidebooks. Our preliminary analysis reveals distinct patterns in LRM reasoning, such as the transition dynamics between cognitive states. This framework provides a theoretically grounded methodology for interpreting LRM cognition and enables future work on more controllable and transparent reasoning systems.","short_abstract":"While Large Reasoning Models (LRMs) generate extensive chain-of-thought reasoning, we lack a principled framework for understanding how these thoughts are structured. In this paper, we introduce a novel approach by applying Schoenfeld's Episode Theory, a classic cognitive framework for human mathematical problem-solvin...","url_abs":"https://arxiv.org/abs/2509.14662","url_pdf":"https://arxiv.org/pdf/2509.14662v1","authors":"[\"Ming Li\",\"Nan Zhang\",\"Chenrui Fan\",\"Hong Jiao\",\"Yanbin Fu\",\"Sydney Peters\",\"Qingshu Xu\",\"Robert Lissitz\",\"Tianyi Zhou\"]","published":"2025-09-18T06:42:41Z","proceeding":"cs.AI","tasks":"[\"cs.AI\",\"cs.CL\",\"cs.LG\"]","methods":"[]","has_code":false}