Mirroring the Mind: Distilling Human-Like Metacognitive Strategies into Large Language Models

Researchers propose Metacognitive Behavioral Tuning (MBT), a new framework that addresses structural fragility in Large Reasoning Models by injecting human-like self-regulatory control into AI thought processes. The approach reduces reasoning collapse and improves accuracy while consuming fewer computational tokens across multi-hop question-answering benchmarks.

• →Large Reasoning Models often fail complex tasks due to poor self-regulatory control rather than lack of reasoning capacity.
• →MBT framework uses two approaches: synthesizing rigorous reasoning traces and rewriting initial traces to stabilize exploration patterns.
• →The method achieves higher accuracy with significantly reduced token consumption compared to baseline models.
• →MBT successfully eliminates reasoning collapse by internalizing metacognitive strategies similar to human thinking.
• →Experiments show consistent outperformance on challenging multi-hop question-answering benchmarks.