pith. sign in

arxiv: 2503.09501 · v3 · pith:5TOP3QL7new · submitted 2025-03-12 · 💻 cs.AI · cs.CL· cs.LG· cs.MA

ReMA: Learning to Meta-think for LLMs with Multi-Agent Reinforcement Learning

classification 💻 cs.AI cs.CLcs.LGcs.MA
keywords reasoningmeta-thinkingremalearningllmsagentagentsfurther
0
0 comments X
read the original abstract

Recent research on Reasoning of Large Language Models (LLMs) has sought to further enhance their performance by integrating meta-thinking -- enabling models to monitor, evaluate, and control their reasoning processes for more adaptive and effective problem-solving. However, current single-agent work lacks a specialized design for acquiring meta-thinking, resulting in low efficacy. To address this challenge, we introduce Reinforced Meta-thinking Agents (ReMA), a novel framework that leverages Multi-Agent Reinforcement Learning (MARL) to elicit meta-thinking behaviors, encouraging LLMs to think about thinking. ReMA decouples the reasoning process into two hierarchical agents: a high-level meta-thinking agent responsible for generating strategic oversight and plans, and a low-level reasoning agent for detailed executions. Through iterative reinforcement learning with aligned objectives, these agents explore and learn collaboration, leading to improved generalization and robustness. Empirical results from single-turn experiments demonstrate that ReMA outperforms single-agent RL baselines on complex reasoning tasks, including competitive-level mathematical benchmarks and LLM-as-a-Judge benchmarks. Additionally, we further extend ReMA to multi-turn interaction settings, leveraging turn-level ratio and parameter sharing to improve efficiency. Comprehensive ablation studies further illustrate the evolving dynamics of each distinct agent, providing valuable insights into how the meta-thinking reasoning process enhances the reasoning capabilities of LLMs. Our code can be found in https://github.com/ziyuwan/ReMA-public

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. HiPO: Hierarchical Preference Optimization for Adaptive Reasoning in LLMs

    cs.AI 2026-04 unverdicted novelty 7.0

    HiPO improves LLM reasoning performance by optimizing preferences separately on response segments rather than entire outputs.

  2. Weak-Link Optimization for Multi-Agent Reasoning and Collaboration

    cs.AI 2026-04 unverdicted novelty 7.0

    WORC improves multi-agent LLM reasoning to 82.2% average accuracy by predicting and compensating for the weakest agent via targeted extra sampling rather than uniform reinforcement.

  3. Where Do CoT Training Gains Land in LLM based Agents?

    cs.AI 2026-06 unverdicted novelty 6.0

    CoT training in LLM agents improves prompt-action quality more than the advantage of generated reasoning, and selectively masking action supervision improves out-of-domain generalization.

  4. Implicit Hierarchical GRPO: Decoupling Tool Invocation from Execution for Tool-Integrated Mathematical Reasoning

    cs.CL 2026-05 unverdicted novelty 6.0

    IH-GRPO introduces implicit hierarchical control via a surrogate loss to decouple tool invocation from execution in LLMs, reporting 1.87-2.53% gains on math reasoning benchmarks.

  5. AIPO: Learning to Reason from Active Interaction

    cs.CL 2026-05 unverdicted novelty 6.0

    AIPO trains LLMs to expand their reasoning capability boundary via active multi-agent interaction with Verify, Knowledge, and Reasoning agents during RLVR, using importance sampling and clipping to handle feedback, th...

  6. AIPO: Learning to Reason from Active Interaction

    cs.CL 2026-05 unverdicted novelty 6.0

    AIPO adds active multi-agent consultation (Verify, Knowledge, Reasoning agents) plus custom importance sampling to RLVR training so LLMs expand their reasoning boundary and then operate without the agents.

  7. Memory-R2: Fair Credit Assignment for Long-Horizon Memory-Augmented LLM Agents

    cs.LG 2026-05 unverdicted novelty 5.0

    Memory-R2 proposes LoGo-GRPO to fix unfair trajectory comparisons in RL training of memory-augmented LLM agents by combining global end-to-end rewards with local rerollouts from identical memory states.

  8. PASK: Toward Intent-Aware Proactive Agents with Long-Term Memory

    cs.AI 2026-04 unverdicted novelty 4.0

    PASK introduces the DD-MM-PAS paradigm for streaming proactive agents with intent-aware detection, hybrid memory modeling, and a new real-world benchmark where the IntentFlow model matches top LLMs on latency while fi...

  9. Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems

    cs.AI 2025-03 unverdicted novelty 2.0

    This survey frames foundation agents using brain-inspired modular architectures and reviews challenges in evolution, collaboration, and safety.