Meta-R1: Empowering Large Reasoning Models with Metacognition

2025/08/24 by Dong, Haonan, Ye, Haoran, Zhu, Wenhao +2 · 1 citation
#Artificial Intelligence (cs.AI) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2508.17291

Abstract

Large Reasoning Models (LRMs) demonstrate remarkable capabilities on complex tasks, exhibiting emergent, human-like thinking patterns. Despite their advances, we identify a fundamental limitation: current LRMs lack a dedicated meta-level cognitive system-an essential faculty in human cognition that enables "thinking about thinking". This absence leaves their emergent abilities uncontrollable (non-adaptive reasoning), unreliable (intermediate error), and inflexible (lack of a clear methodology). To address this gap, we introduce Meta-R1, a systematic and generic framework that endows LRMs with explicit metacognitive capabilities. Drawing on principles from cognitive science, Meta-R1 decomposes the reasoning process into distinct object-level and meta-level components, orchestrating proactive planning, online regulation, and adaptive early stopping within a cascaded framework. Experiments on three challenging benchmarks and against eight competitive baselines demonstrate that Meta-R1 is: (I) high-performing, surpassing state-of-the-art methods by up to 27.3%; (II) token-efficient, reducing token consumption to 15.7% ~ 32.7% and improving efficiency by up to 14.8% when compared to its vanilla counterparts; and (III) transferable, maintaining robust performance across datasets and model backbones.

Citations

• MemOS: A Memory OS for AI System
• Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency
• Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning
• Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space
• AdaptThink: Reasoning Models Can Learn When to Think
• LLM-BABYBENCH: Understanding and Evaluating Grounded Planning and Reasoning in LLMs
• S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models
• Dynamic Early Exit in Reasoning Models
• Generative AI Act II: Test Time Scaling Drives Cognition Engineering
• Reasoning Models Can Be Effective Without Thinking
• Speculative Thinking: Enhancing Small-Model Reasoning with Large Model Guidance at Inference Time
• Continual Learning of Multiple Cognitive Functions with Brain-inspired Temporal Development Mechanism
• MegaMath: Pushing the Limits of Open Math Corpora
• DeepSeek-R1 Thoughtology: Let's think about LLM Reasoning
• A Survey of Efficient Reasoning for Large Reasoning Models: Language, Multimodality, and Beyond
• I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders
• MetaScale: Test-Time Scaling with Evolving Meta-Thoughts
• Instruct-of-Reflection: Enhancing Large Language Models Iterative Reflection Capabilities via Dynamic-Meta Instruction
• Can Large Language Models Detect Errors in Long Chain-of-Thought Reasoning?
• Chain of Draft: Thinking Faster by Writing Less
• Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning
• LightThinker: Thinking Step-by-Step Compression
• SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMs
• TokenSkip: Controllable Chain-of-Thought Compression in LLMs
• Probing Semantic Routing in Large Mixture-of-Expert Models
• CoT-Valve: Length-Compressible Chain-of-Thought Tuning
• Training Language Models to Reason Efficiently
• LIMO: Less is More for Reasoning
• Demystifying Long Chain-of-Thought Reasoning in LLMs
• DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning
• rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking
• Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs
• Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoning
• Reasoning Beyond Words ? Exploring framework for hidden state reasoning
• Non-myopic Generation of Language Models for Reasoning and Planning
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters
• Memory³: Language Modeling with Explicit Memory
• Dual Reasoning: A GNN-LLM Collaborative Framework for Knowledge Graph Question Answering
• BioXP-0.5B: Explainable Medical-AI via RL-GRPO
• ReEvo: Large Language Models as Hyper-Heuristics with Reflective Evolution
• Unveiling Theory of Mind in Large Language Models: A Parallel to Single Neurons in the Human Brain
• Training Verifiers to Solve Math Word Problems
• Measuring Mathematical Problem Solving With the MATH Dataset
• Limitations of Autoregressive Models and Their Alternatives
• Moving from Novice to Expertise and Its Implications for Instruction
• A BEHAVIORAL MODEL OF RATIONAL CHOICE

Cited by

• Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning

Related

• Cognitive Foundations for Reasoning and Their Manifestation in LLMs
• From "Aha Moments" to Controllable Thinking: Toward Meta-Cognitive Reasoning in Large Reasoning Models via Decoupled Reasoning and Control
• CoRE: Enhancing Metacognition with Label-free Self-evaluation in LRMs
• MR-Ben: A Meta-Reasoning Benchmark for Evaluating System-2 Thinking in LLMs
• Meta Reasoning for Large Language Models