mlsanityData to 14 Oct 2026

ThinkSwitcher: When to Think Hard, When to Think Fast

arXiv 2505.14183 ↗First posted 20 May 2025cs.CLcited by 37

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

Abstract (arXiv)

Large reasoning models (LRMs) excel at solving complex tasks by leveraging long chain-of-thought (CoT) reasoning. However, this often leads to overthinking on simple tasks, resulting in unnecessary computational overhead. We observe that LRMs inherently possess the capability for efficient short CoT reasoning, which can be reliably elicited through prompt design. To leverage this capability, we propose ThinkSwitcher, a framework that enables a single LRM to dynamically switch between short and long CoT modes based on task complexity. ThinkSwitcher introduces a lightweight switching module trained with supervision signals derived from the relative performance of each reasoning mode across tasks. Experiments on multiple reasoning benchmarks show that ThinkSwitcher reduces computational cost by 20-30% while maintaining high accuracy on complex tasks. This demonstrates the effectiveness of ThinkSwitcher as a scalable and efficient solution for unified LRM deployment.

Key facts. Cited by 37 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 26 of its 49 references that are in the map. First posted on arXiv 20 May 2025.

Builds on

26 of 49 references.

Closest ancestors · 10

Same direction, newest first

  1. 2025Llama-Nemotron: Efficient Reasoning Models2 May
  2. 2025Reasoning Models Can Be Effective Without Thinking14 Apr
  3. 2025A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond27 Mar
  4. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models20 Mar
  5. 2025DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models6 Mar
  6. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
  7. 2025The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks12 Feb
  8. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruning22 Jan
  9. 2024Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs30 Dec
  10. 2024C3oT: Generating Shorter Chain-of-Thought without Compromising Effectiveness16 Dec

Other references · 16

  1. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  3. 2025Qwen3 Technical Reportcited by 8,820
  4. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  5. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  6. 2024OpenAI o1 System Cardcited by 1,244
  7. 2025Kimi k1.5: Scaling Reinforcement Learning with LLMscited by 1,210
  8. 2024Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inferencecited by 554
  9. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 511
  10. 2018Tune: A Research Platform for Distributed Model Selection and Trainingcited by 395
  11. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 377
  12. 2025From System 1 to System 2: A Survey of Reasoning Large Language Modelscited by 320
4 more
  1. 2025Chain of Draft: Thinking Faster by Writing Lesscited by 297
  2. 2025OverThink: Slowdown Attacks on Reasoning LLMscited by 96
  3. 2018Deep Learning using Rectified Linear Units (ReLU)cited by 48
  4. 2025Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluationcited by 29

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 17; 2026: 20.

37 papers.

Sort:

By area:

  1. 2025Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning Modelscited by 53
  2. 2025Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Modelscited by 37
  3. 2025Towards Concise and Adaptive Thinking in Large Reasoning Models: A Surveycited by 18
  4. 2025Aware First, Think Less: Dynamic Boundary Self-Awareness Drives Extreme Reasoning Efficiency in Large Language Modelscited by 13
  5. 2025From Efficiency to Adaptivity: A Deeper Look at Adaptive Reasoning in Large Language Modelscited by 12
  6. 2025DiffAdapt: Difficulty-Adaptive Reasoning for Token-Efficient LLM Inferencecited by 11
  7. 2025How Far Are We from Optimal Reasoning Efficiency?cited by 8
  8. 2025Hierarchical Budget Policy Optimization for Adaptive Reasoningcited by 8
  9. 2025Test-Time Scaling in Diffusion LLMs via Hidden Semi-Autoregressive Expertscited by 8
  10. 2026FoE: Forest of Errors Makes the First Solution the Best in Large Reasoning Modelscited by 8
  11. 2025Asynchronous Reasoning: Training-Free Interactive Thinking LLMscited by 7
  12. 2025Scaling Up, Speeding Up: A Benchmark of Speculative Decoding for Efficient LLM Test-Time Scalingcited by 6
25 more
  1. 2025On the Provable Performance Guarantee of Efficient Reasoning Modelscited by 6
  2. 2026MC-GRPO: Median-Centered Group Relative Policy Optimization for Small-Rollout Reinforcement Learningcited by 6
  3. 2025Putting on the Thinking Hats: A Survey on Chain of Thought Fine-tuning from the Perspective of Human Reasoning Mechanismcited by 4
  4. 2025The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Modelscited by 4
  5. 2026Stable Adaptive Thinking via Advantage Shaping and Length-Aware Gradient Regulationcited by 4
  6. 2026Compress the Easy, Explore the Hard: Difficulty-Aware Entropy Regularization for Efficient LLM Reasoningcited by 4
  7. 2026Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewardscited by 3
  8. 2025Fast, Slow, and Tool-augmented Thinking for LLMs: A Reviewcited by 2
  9. 2025Adaptive Dual Reasoner: Large Reasoning Models Can Think Efficiently by Hybrid Reasoningcited by 2
  10. 2026PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoningcited by 2
  11. 2026SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinkingcited by 2
  12. 2026Reasoning Is Not Free: Robust Adaptive Cost-Efficient Routing for LLM-as-a-Judgecited by 2
  13. 2025SCALE: Selective Resource Allocation for Overcoming Performance Bottlenecks in Mathematical Test-time Scalingcited by 1
  14. 2026Anytime Safe PAC Efficient Reasoningcited by 1
  15. 2026VisRef: Visual Refocusing while Thinking Improves Test-Time Scaling in Multi-Modal Large Reasoning Modelscited by 1
  16. 2026EvoThink: Evolving Thinking in Large Reasoning Models via Self-Pruning and Aha-Moment Preference Optimizationcited by 1
  17. 2026APR: Penalizing Structural Redundancy in Large Reasoning Models via Anchor-based Process Rewardscited by 0
  18. 2026Post-Reasoning: Improving the Performance of Non-Thinking Models at No Costcited by 0
  19. 2026Nice Fold or Hero Call: Learning Budget-Efficient Thinking for Adaptive Reasoningcited by 0
  20. 2026HRBench: Benchmarking and Understanding Thinking-Mode Switch Strategies in Hybrid-Reasoning LLMscited by 0
  21. 2026REPOT: Recoverable Program-of-Thought via Checkpoint Repaircited by 0
  22. 2026Overthink-Triggered Slowdown Attacks on LVLM-Based Robotic Systemscited by 0
  23. 2026ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scalingcited by 0
  24. 2026MetaCtrl: Your Large Language Models Can Reason Better and More Concisely with a Metacognitive Controllercited by 0
  25. 2026HARISSA: Inference-Time Self-Checks for Efficient and Safe Local Language Model Deploymentcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Long or short CoT? Investigating Instance-level Switch of Large Reasoning Models
  2. 2025Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Models
  3. 2026CoTJudger: A Graph-Driven Framework for Automatic Evaluation of Chain-of-Thought Efficiency and Redundancy in LRMs
  4. 2025THINK-Bench: Evaluating Thinking Efficiency and Chain-of-Thought Quality of Large Reasoning Models
  5. 2026TERMINATOR: Learning Optimal Exit Points for Early Stopping in Chain-of-Thought Reasoning
  6. 2025TrimR: Verifier-based Training-Free Thinking Compression for Efficient Test-Time Scaling
  7. 2025Reasoning Efficiently Through Adaptive Chain-of-Thought Compression: A Self-Optimizing Framework
  8. 2025ThinkPilot: Steering Reasoning Models via Automated Think-prefixes Optimization