When to Continue Thinking: Adaptive Thinking Mode Switching for Efficient Reasoning
Authors:
- Xiaoyun ZhangXZXiaoyun Zhang
- Jingqing RuanJRJingqing Ruan
- Xing MaXMXing Ma
- Yawen ZhuYZYawen Zhu
- Haodong ZhaoHZHaodong Zhao
- Hao LiHLHao Li
- Jiansong ChenJCJiansong Chen
- Ke ZengKZKe Zeng
- Xunliang CaiXCXunliang Cai
Preprint
Abstract
Large reasoning models (LRMs) achieve remarkable performance via long reasoning chains, but often incur excessive computational overhead due to redundant reasoning, especially on simple tasks. In this work, we systematically quantify the upper bounds of LRMs under both Long-Thinking and No-Thinking modes, and uncover the phenomenon of "Internal Self-Recovery Mechanism" where models implicitly supplement reasoning during answer generation. Building on this insight, we propose Adaptive Self-Recovery Reasoning (ASRR), a framework that suppresses unnecessary reasoning and enables implicit recovery. By introducing accuracy-aware length reward regulation, ASRR adaptively allocates reasoning effort according to problem difficulty, achieving high efficiency with negligible performance sacrifice. Experiments across multiple benchmarks and models show that, compared with GRPO, ASRR reduces reasoning budget by up to 32.5% (1.5B) and 25.7% (7B) with minimal accuracy loss (1.2% and 0.6% pass@1), and significantly boosts harmless rates on safety benchmarks (up to +21.7%). Our results highlight the potential of ASRR for enabling efficient, adaptive, and safer reasoning in LRMs.Keywords
All Related Versions
This publication has 9 references indexed in Scilit:
- DAST: Difficulty-Adaptive Slow-Thinking for Large Reasoning Models2025
- Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?2025
- SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs2025
- Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought2024
- Direct Preference Optimization: Your Language Model is Secretly a Reward Model2023
- Tree of Thoughts: Deliberate Problem Solving with Large Language Models2023
- BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Dataset2023
- Self-Evaluation Guided Beam Search for Reasoning2023
- Chain-Of-Thought Prompting Elicits Reasoning in Large Language Models2022