mlsanityData to 14 Oct 2026

ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization

arXiv 2510.12063 ↗First posted 14 Oct 2025cs.AI cs.CLcited by 3

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

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

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 49 of its 59 references that are in the map. First posted on arXiv 14 Oct 2025.

Builds on

49 of 59 references.

Closest ancestors · 21

Same direction, newest first

  1. 2025Understanding Reasoning in Thinking Language Models via Steering Vectors22 Jun
  2. 2025Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency10 Jun
  3. 2025Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Models4 Jun
  4. 2025Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning3 Jun
  5. 2025Not All Tokens Are What You Need In Thinking23 May
  6. 2025Between Underthinking and Overthinking: An Empirical Study of Reasoning Length and correctness in LLMs30 Apr
  7. 2025Dynamic Early Exit in Reasoning Models22 Apr
  8. 2025Sleep-time Compute: Beyond Inference Scaling at Test-time17 Apr
  9. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  10. 2025Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification7 Apr
  11. 2025Critical Thinking: Which Kinds of Complexity Govern Optimal Reasoning Length?2 Apr
  12. 2025ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning2 Apr
9 more
  1. 2025Effectively Controlling Reasoning Models through Thinking Intervention31 Mar
  2. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  3. 2025L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning6 Mar
  4. 2025CoT-Valve: Length-Compressible Chain-of-Thought Tuning13 Feb
  5. 2025The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks12 Feb
  6. 2025Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning26 Jan
  7. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  8. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  9. 2024Efficiently Scaling LLM Reasoning with Certaindex30 Dec

Other references · 28

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  3. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  4. 2024Qwen2.5 Technical Reportcited by 4,942
  5. 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,807
  6. 2024OpenAI o1 System Cardcited by 1,244
  7. 2023Instruction-Following Evaluation for Large Language Modelscited by 1,228
  8. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 832
  9. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
  10. 2023Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolutioncited by 551
  11. 2023XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Modelscited by 505
  12. 2025Reasoning Models Don't Always Say What They Thinkcited by 415
16 more
  1. 2024A StrongREJECT for Empty Jailbreakscited by 385
  2. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 384
  3. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 297
  4. 2025Training Language Models to Reason Efficientlycited by 247
  5. 2023EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizerscited by 193
  6. 2024Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understandingcited by 188
  7. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 183
  8. 2025MultiChallenge: A Realistic Multi-Turn Conversation Evaluation Benchmark Challenging to Frontier LLMscited by 180
  9. 2025SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilitiescited by 136
  10. 2025Thought Anchors: Which LLM Reasoning Steps Matter?cited by 129
  11. 2023Thread of Thought Unraveling Chaotic Contextscited by 72
  12. 2025RealSafe-R1: Safety-Aligned DeepSeek-R1 without Compromising Reasoning Capabilitycited by 59
  13. 2025No Free Lunch: Rethinking Internal Feedback for LLM Reasoningcited by 37
  14. 2023Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logiccited by 35
  15. 2025Understanding the Thinking Process of Reasoning Models: A Perspective from Schoenfeld's Episode Theorycited by 9
  16. 2023Chain-of-Symbol Prompting Elicits Planning in Large Langauge Modelscited by 6

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Co-FactChecker: A Framework for Human-AI Collaborative Claim Verification Using Large Reasoning Modelscited by 1
  2. 2026Aligning Large Vision-Language Models at Test Time: A Trajectory-Guided Structured Sampling Approachcited by 1
  3. 2026DyCon: Dynamic Reasoning Control via Evolving Difficulty Modelingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Don't Think Longer, Think Wisely: Optimizing Thinking Dynamics for Large Reasoning Models
  2. 2025ThinkSwitcher: When to Think Hard, When to Think Fast
  3. 2025Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time
  4. 2025Efficient Inference for Large Reasoning Models: A Survey
  5. 2026InjectRBP: Steering Large Language Model Reasoning Behavior via Pattern Injection
  6. 2025Implicit Reasoning in Large Language Models: A Comprehensive Survey
  7. 2025Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models
  8. 2025Efficient Reasoning Models: A Survey