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

2026-02-25

AI Summary

This work addresses the tendency of large reasoning models to suffer from reasoning collapse in complex tasks due to insufficient self-regulation, often failing to reach correct final answers despite accurate intermediate steps. To mitigate this, the authors propose a Metacognitive Behavior Tuning (MBT) framework that systematically integrates human-inspired metacognitive mechanisms into large language models through post-training behavioral fine-tuning, enabling models to internalize self-monitoring and regulation strategies. MBT comprises two complementary approaches: synthesizing reasoning trajectories from scratch (MBT-S) and rewriting the model’s initial reasoning attempts (MBT-R). Evaluated on multi-hop question answering benchmarks, MBT significantly outperforms baseline methods, achieving higher accuracy while substantially reducing reasoning token consumption, thereby enabling more stable and efficient inference.

Technology Category

• Search and Optimization: Metareasoning and Metaheuristics
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large Reasoning Models (LRMs) often exhibit structural fragility in complex reasoning tasks, failing to produce correct answers even after successfully deriving valid intermediate steps. Through systematic analysis, we observe that these failures frequently stem not from a lack of reasoning capacity, but from a deficiency in self-regulatory control, where valid logic is destabilized by uncontrolled exploration or the failure to recognize logical sufficiency. Motivated by this observation, we propose Metacognitive Behavioral Tuning (MBT), a post-training framework that explicitly injects metacognitive behaviors into the model's thought process. MBT implements this via two complementary formulations: (1) MBT-S, which synthesizes rigorous reasoning traces from scratch, and (2) MBT-R, which rewrites the student's initial traces to stabilize intrinsic exploration patterns. Experiments across multi-hop QA benchmarks demonstrate that MBT consistently outperforms baselines, achieving notable gains on challenging benchmarks. By effectively eliminating reasoning collapse, MBT achieves higher accuracy with significantly reduced token consumption, demonstrating that internalizing metacognitive strategies leads to more stable and robust reasoning.

Problem

• Large Reasoning Models
• metacognition
• self-regulatory control
• reasoning collapse
• complex reasoning tasks

Innovation

• Metacognitive Behavioral Tuning
• reasoning collapse
• Large Reasoning Models
• self-regulatory control
• reasoning trace synthesis