Jian Luan

Guanzhong Chen, Shaoxiong Yang, Chao Li et al.

Large Language Models (LLMs) have achieved remarkable success across diverse natural language processing tasks, yet their deployment in real-world applications is hindered by fixed knowledge cutoffs and difficulties in generating controllable, accurate outputs in a single inference. Multi-agent systems (MAS) built from specialized LLM agents offer a promising solution, enabling dynamic collaboration and iterative reasoning. However, optimizing these systems remains a challenge, as conventional methods such as prompt engineering and supervised fine-tuning entail high engineering overhead and limited adaptability. Reinforcement learning (RL), particularly multi-agent reinforcement learning (MARL), provides a scalable framework by refining agent policies based on system-level feedback. Nevertheless, existing MARL algorithms, such as Multi-Agent Proximal Policy Optimization (MAPPO), rely on Critic networks, which can cause training instability and increase computational burden. To address these limitations and target the prototypical Multi-Agent Search System (MASS), we propose Multi-Agent Heterogeneous Group Policy Optimization (MHGPO), a novel Critic-free algorithm that guides policy updates by estimating relative reward advantages across heterogeneous groups of rollouts. MHGPO eliminates the need for Critic networks, enhancing stability and reducing computational overhead. Additionally, we introduce three group rollout sampling strategies that trade off between efficiency and effectiveness. Experiments on a multi-agent LLM-based search system demonstrate that MHGPO consistently outperforms MAPPO in both task performance and computational efficiency, without requiring warm-up, underscoring its potential for stable and scalable optimization of complex LLM-based MAS.

Jinpeng Wang, Chao Li, Ting Ye et al.

Reinforcement Learning with Verifiable Rewards (RLVR) demonstrates significant potential in enhancing the reasoning capabilities of Large Language Models (LLMs). However, existing RLVR methods are often constrained by issues such as coarse-grained rewards, reward noise, and inefficient exploration, which lead to unstable training and entropy collapse. To address this challenge, we propose the Intrinsic Confidence-Driven Group Relative Preference Optimization method (ICPO). The intuition behind it lies in the fact that the probabilities of an LLM generating different responses can inherently and directly reflect its self-assessment of the reasoning process. Inspired by the idea of preference modeling, ICPO calculates a preference advantage score for each response by comparing the relative generation probabilities of multiple responses under the same input prompt, and integrates this score with verifiable rewards to guide the exploration process. We have discovered that the preference advantage score not only alleviates the issues of coarse-grained rewards and reward noise but also effectively curbs overconfident errors, enhances the relative superiority of undervalued high-quality responses, and prevents the model from overfitting to specific strategies. Comprehensive experiments across four general-domain benchmarks and three mathematical benchmarks demonstrate that ICPO steadily boosts reasoning compared to GRPO.