AdaptThink: Reasoning Models Can Learn When to Think

• URL: http://arxiv.org/abs/2505.13417v1
• Date: Mon, 19 May 2025 17:50:52 GMT
• Title: AdaptThink: Reasoning Models Can Learn When to Think
• Authors: Jiajie Zhang, Nianyi Lin, Lei Hou, Ling Feng, Juanzi Li,
• Abstract summary: We propose AdaptThink, a novel RL algorithm to teach reasoning models to choose the optimal thinking mode adaptively based on problem difficulty.<n>Our experiments indicate that AdaptThink significantly reduces the inference costs while further enhancing performance.
• Score: 42.77877234302026
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recently, large reasoning models have achieved impressive performance on various tasks by employing human-like deep thinking. However, the lengthy thinking process substantially increases inference overhead, making efficiency a critical bottleneck. In this work, we first demonstrate that NoThinking, which prompts the reasoning model to skip thinking and directly generate the final solution, is a better choice for relatively simple tasks in terms of both performance and efficiency. Motivated by this, we propose AdaptThink, a novel RL algorithm to teach reasoning models to choose the optimal thinking mode adaptively based on problem difficulty. Specifically, AdaptThink features two core components: (1) a constrained optimization objective that encourages the model to choose NoThinking while maintaining the overall performance; (2) an importance sampling strategy that balances Thinking and NoThinking samples during on-policy training, thereby enabling cold start and allowing the model to explore and exploit both thinking modes throughout the training process. Our experiments indicate that AdaptThink significantly reduces the inference costs while further enhancing performance. Notably, on three math datasets, AdaptThink reduces the average response length of DeepSeek-R1-Distill-Qwen-1.5B by 53% and improves its accuracy by 2.4%, highlighting the promise of adaptive thinking-mode selection for optimizing the balance between reasoning quality and efficiency. Our codes and models are available at https://github.com/THU-KEG/AdaptThink.

Related papers

• Thinking with Nothinking Calibration: A New In-Context Learning Paradigm in Reasoning Large Language Models [23.642200042199484]
  We propose Thinking with Nothinking (JointThinking) as an in-context learning (ICL) paradigm for Reasoning large language models (RLLMs)<n>Our method prompts the model to generate two answers in parallel: one in Thinking mode and the other in Nothinking mode.<n>JointThinking significantly outperforms few-shot chain-of-thought robustness (CoT) and majority voting with improved answer.
  arXiv  Detail & Related papers  (2025-08-05T12:09:55Z)
• Does Thinking More always Help? Understanding Test-Time Scaling in Reasoning Models [103.03315678501546]
  Extending thinking traces using prompts like "Wait" or "Let me rethink" can improve performance.<n>This raises a natural question: Does thinking more at test-time truly lead to better reasoning?<n>We show a consistent pattern of initial performance improvements from additional thinking followed by a decline, due to "overthinking"
  arXiv  Detail & Related papers  (2025-06-04T17:55:09Z)
• Don't Think Longer, Think Wisely: Optimizing Thinking Dynamics for Large Reasoning Models [68.96619605651155]
  Large reasoning models (LRMs) may drastically increase the output length due to overthinking.<n>We propose a dynamic optimization framework that segments model-generated reasoning paths into distinct thinking patterns.<n>Our method achieves up to a 12% accuracy improvement and reducing token usage from approximately 5,000 to 3,000 tokens.
  arXiv  Detail & Related papers  (2025-05-27T20:59:29Z)
• Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Models [45.33952788910874]
  TON is a two-stage training strategy for vision-language models.<n>It introduces a think-or-not format that serves as a cold start for selective reasoning.<n>TON can reduce the completion length by up to 90% compared to vanilla GRPO.
  arXiv  Detail & Related papers  (2025-05-22T16:13:29Z)
• Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning [75.04643265875072]
  Large reasoning models (LRMs) have demonstrated strong performance on complex reasoning tasks, but often suffer from overthinking.<n>Inspired by the dual process theory in cognitive science, we propose Adaptive Cognition Policy Optimization.<n>ACPO enables LRMs to achieve efficient reasoning through adaptive cognitive allocation and dynamic system switch.
  arXiv  Detail & Related papers  (2025-05-22T07:15:08Z)
• Thinkless: LLM Learns When to Think [57.857534644932194]
  Reasoning Language Models, capable of extended chain-of-thought reasoning, have demonstrated remarkable performance on tasks requiring complex logical inference.<n>We propose Thinkless, a learnable framework that empowers an LLM to adaptively select between short-form and long-form reasoning.<n>On several benchmarks such as Minerva Algebra, MATH-500, and GSM8K, Thinkless is able to reduce the usage of long-chain thinking by 50% - 90%.
  arXiv  Detail & Related papers  (2025-05-19T17:24:16Z)
• AdaR1: From Long-CoT to Hybrid-CoT via Bi-Level Adaptive Reasoning Optimization [86.56120216550232]
  We propose a novel two-stage framework for adaptive and efficient reasoning.<n>First, we construct a hybrid reasoning model by merging long and short CoT models.<n>Second, we apply bi-level preference training to guide the model to select suitable reasoning styles.
  arXiv  Detail & Related papers  (2025-04-30T14:01:45Z)
• VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning [55.97950660659051]
  We aim to enhance the slow-thinking capabilities of vision-language models using reinforcement learning (without relying on distillation)<n>We introduce Forced Rethinking, which appends a rethinking trigger token to the end of rollouts in RL training, explicitly enforcing a self-reflection reasoning step.<n>Our model, VL-Rethinker, advances state-of-the-art scores on MathVista, MathVerse to achieve 80.4%, 63.5% respectively.
  arXiv  Detail & Related papers  (2025-04-10T17:41:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.