{"ID":2855485,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.12063","arxiv_id":"2510.12063","title":"ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization","abstract":"Large Reasoning Models (LRMs) are powerful, but they still suffer from inefficient and off-target reasoning. Currently, training-free methods are limited to either rigid heuristics or descriptive, non-actionable analyses. In this paper, we introduce ThinkPilot, a training-free framework that automatically optimizes LRMs reasoning. It uses an evolutionary process to generate think-prefixes, which are instructions that evolve driven by a taxonomy of reasoning behaviors to guide models toward superior performance. Extensive experiments demonstrate ThinkPilot's broad effectiveness: it significantly improves the accuracy-length trade-off for efficient reasoning, drastically improves safety (for example, cutting the StrongREJECT score of DeepSeek-R1-Distill-Qwen-32B from 27.0% to 0.7), and enhances instruction following. It also synergizes with existing training-based methods. Our analysis reveals that think-prefixes can reliably control LRMs' reasoning behaviors, and that different tasks have strong preferences for specific behavioral distributions. By automatically identifying and eliciting these behaviors, ThinkPilot provides a generalizable framework for aligning LRMs reasoning with task demands. Data and code are available at https://github.com/teqkilla/ThinkPilot","short_abstract":"Large Reasoning Models (LRMs) are powerful, but they still suffer from inefficient and off-target reasoning. Currently, training-free methods are limited to either rigid heuristics or descriptive, non-actionable analyses. In this paper, we introduce ThinkPilot, a training-free framework that automatically optimizes LRM...","url_abs":"https://arxiv.org/abs/2510.12063","url_pdf":"https://arxiv.org/pdf/2510.12063v1","authors":"[\"Sunzhu Li\",\"Zhiyu Lin\",\"Shuling Yang\",\"Jiale Zhao\",\"Wei Chen\"]","published":"2025-10-14T02:02:19Z","proceeding":"cs.AI","tasks":"[\"cs.AI\",\"cs.CL\"]","methods":"[]","has_code":false,"code_links":[{"ID":608252,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2855485,"paper_url":"https://arxiv.org/abs/2510.12063","paper_title":"ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization","repo_url":"https://github.com/teqkilla/ThinkPilot","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}