Mitigating Overthinking in Large Reasoning Language Models via Reasoning Path Deviation Monitoring

Researchers propose a new early-exit method for Large Reasoning Language Models that detects and prevents overthinking by monitoring high-entropy transition tokens that indicate deviation from correct reasoning paths. The method improves performance and efficiency compared to existing approaches without requiring additional training overhead or limiting inference throughput.

• →Large Reasoning Language Models often suffer from overthinking, generating redundant reasoning steps that hurt performance and efficiency.
• →Current early-exit strategies to combat overthinking either require extra training or reduce inference speed.
• →The new method uses path deviation monitoring to detect high-entropy transition tokens that signal reasoning errors.
• →The approach is integrated into the native reasoning process without requiring proxy models or content switching.
• →Experiments show the method delivers the largest performance improvement over vanilla Chain-of-Thought reasoning compared to existing solutions.