Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning

• URL: http://arxiv.org/abs/2509.03817v1
• Date: Thu, 04 Sep 2025 02:06:06 GMT
• Title: Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning
• Authors: Wei Yang, Jesse Thomason,
• Abstract summary: We introduce the Meta-Policy Deliberation Framework (MPDF), where agents learn a decentralized policy over a set of high-level meta-cognitive actions.<n>We develop SoftRankPO, a novel reinforcement learning algorithm.
• Score: 13.9097452260769
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multi-agent systems of large language models (LLMs) show promise for complex reasoning, but their effectiveness is often limited by fixed collaboration protocols. These frameworks typically focus on macro-level orchestration while overlooking agents' internal deliberative capabilities. This critical meta-cognitive blindspot treats agents as passive executors unable to adapt their strategy based on internal cognitive states like uncertainty or confidence. We introduce the Meta-Policy Deliberation Framework (MPDF), where agents learn a decentralized policy over a set of high-level meta-cognitive actions: Persist, Refine, and Concede. To overcome the instability of traditional policy gradients in this setting, we develop SoftRankPO, a novel reinforcement learning algorithm. SoftRankPO stabilizes training by shaping advantages based on the rank of rewards mapped through smooth normal quantiles, making the learning process robust to reward variance. Experiments show that MPDF with SoftRankPO achieves a a 4-5% absolute gain in average accuracy across five mathematical and general reasoning benchmarks compared to six state-of-the-art heuristic and learning-based multi-agent reasoning algorithms. Our work presents a paradigm for learning adaptive, meta-cognitive policies for multi-agent LLM systems, shifting the focus from designing fixed protocols to learning dynamic, deliberative strategies.

Related papers

• Policy-Conditioned Policies for Multi-Agent Task Solving [53.67744322553693]
  In this work, we propose a paradigm shift that bridges the gap by representing policies as human-interpretable source code.<n>We reformulate the learning problem by utilizing Large Language Models (LLMs) as approximate interpreters.<n>We formalize this process as textitProgrammatic Iterated Best Response (PIBR), an algorithm where the policy code is optimized by textual gradients.
  arXiv  Detail & Related papers  (2025-12-24T07:42:10Z)
• MAGIC-MASK: Multi-Agent Guided Inter-Agent Collaboration with Mask-Based Explainability for Reinforcement Learning [0.0]
  We propose a mathematically grounded framework, MAGIC-MASK, that extends perturbation-based explanation to Multi-Agent Reinforcement Learning.<n>Our method integrates Proximal Policy Optimization, adaptive epsilon-greedy exploration, and lightweight inter-agent collaboration.<n>This collaboration enables each agent to perform saliency-guided masking and share reward-based insights with peers, reducing the time required for critical state discovery.
  arXiv  Detail & Related papers  (2025-09-30T20:53:28Z)
• Hierarchical Reinforcement Learning with Low-Level MPC for Multi-Agent Control [1.5856188608650232]
  We propose a hierarchical framework that combines tactical decision-making via reinforcement learning (RL) with low-level execution through Model Predictive Control (MPC)<n>tested on a predator-prey benchmark, our approach outperforms end-to-end and shielding-based RL baselines in terms of reward, safety, and consistency.
  arXiv  Detail & Related papers  (2025-09-19T09:27:15Z)
• The Landscape of Agentic Reinforcement Learning for LLMs: A Survey [103.32591749156416]
  The emergence of agentic reinforcement learning (Agentic RL) marks a paradigm shift from conventional reinforcement learning applied to large language models (LLM RL)<n>This survey formalizes this conceptual shift by contrasting the degenerate single-step Markov Decision Processes (MDPs) of LLM-RL with the temporally extended, partially observable Markov decision processes (POMDPs) that define Agentic RL.
  arXiv  Detail & Related papers  (2025-09-02T17:46:26Z)
• Adversarial Testing in LLMs: Insights into Decision-Making Vulnerabilities [5.0778942095543576]
  This paper introduces an adversarial evaluation framework designed to systematically stress-test the decision-making processes of Large Language Models.<n>We apply this framework to several state-of-the-art LLMs, including GPT-3.5, GPT-4, Gemini-1.5, and DeepSeek-V3.<n>Our findings highlight distinct behavioral patterns across models and emphasize the importance of adaptability and fairness recognition for trustworthy AI deployment.
  arXiv  Detail & Related papers  (2025-05-19T14:50:44Z)
• Collab: Controlled Decoding using Mixture of Agents for LLM Alignment [90.6117569025754]
  Reinforcement learning from human feedback has emerged as an effective technique to align Large Language models.<n>Controlled Decoding provides a mechanism for aligning a model at inference time without retraining.<n>We propose a mixture of agent-based decoding strategies leveraging the existing off-the-shelf aligned LLM policies.
  arXiv  Detail & Related papers  (2025-03-27T17:34:25Z)
• ReMA: Learning to Meta-think for LLMs with Multi-Agent Reinforcement Learning [53.817538122688944]
  We introduce Reinforced Meta-thinking Agents (ReMA) to elicit meta-thinking behaviors from Reasoning of Large Language Models (LLMs)<n>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.<n> Empirical results from single-turn experiments demonstrate that ReMA outperforms single-agent RL baselines on complex reasoning tasks.
  arXiv  Detail & Related papers  (2025-03-12T16:05:31Z)
• From Novice to Expert: LLM Agent Policy Optimization via Step-wise Reinforcement Learning [62.54484062185869]
  We introduce StepAgent, which utilizes step-wise reward to optimize the agent's reinforcement learning process.<n>We propose implicit-reward and inverse reinforcement learning techniques to facilitate agent reflection and policy adjustment.
  arXiv  Detail & Related papers  (2024-11-06T10:35:11Z)
• Strategy Synthesis in Markov Decision Processes Under Limited Sampling Access [3.441021278275805]
  In environments modeled by gray-box Markov decision processes (MDPs), the impact of the agents' actions are known in terms of successor states but not the synthesiss involved. In this paper, we devise a strategy algorithm for gray-box MDPs via reinforcement learning that utilizes interval MDPs as internal model.
  arXiv  Detail & Related papers  (2023-03-22T16:58:44Z)
• FACMAC: Factored Multi-Agent Centralised Policy Gradients [103.30380537282517]
  We propose FACtored Multi-Agent Centralised policy gradients (FACMAC) It is a new method for cooperative multi-agent reinforcement learning in both discrete and continuous action spaces. We evaluate FACMAC on variants of the multi-agent particle environments, a novel multi-agent MuJoCo benchmark, and a challenging set of StarCraft II micromanagement tasks.
  arXiv  Detail & Related papers  (2020-03-14T21:29:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.