MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Yuan, Huining, Xu, Zelai, Tan, Zheyue, Yi, Xiangmin, Guang, Mo, Long, Kaiwen, Hui, Haojia, Li, Boxun, Chen, Xinlei, Zhao, Bo, Zhang, Xiao-Ping, Yu, Chao, Wang, Yu
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912899848470528
author Yuan, Huining
Xu, Zelai
Tan, Zheyue
Yi, Xiangmin
Guang, Mo
Long, Kaiwen
Hui, Haojia
Li, Boxun
Chen, Xinlei
Zhao, Bo
Zhang, Xiao-Ping
Yu, Chao
Wang, Yu
author_facet Yuan, Huining
Xu, Zelai
Tan, Zheyue
Yi, Xiangmin
Guang, Mo
Long, Kaiwen
Hui, Haojia
Li, Boxun
Chen, Xinlei
Zhao, Bo
Zhang, Xiao-Ping
Yu, Chao
Wang, Yu
contents Developing Large Language Models (LLMs) to cooperate and compete effectively within multi-agent systems (MASs) is a critical step towards more advanced intelligence. While reinforcement learning (RL) has proven effective for enhancing reasoning in single-agent tasks, its extension to multi-turn, multi-agent scenarios remains underexplored due to the challenges of long-horizon credit assignment and agent-specific advantage estimation. To address these challenges, we introduce MARSHAL, an end-to-end RL framework that incentivizes Multi-Agent Reasoning through Self-play witH strAtegic LLMs in both cooperative and competitive games. MARSHAL features a turn-level advantage estimator that aligns learning signals with each interaction for credit assignment, and an agent-specific advantage normalization to stabilize multi-agent training. By learning with self-play across cooperative and competitive games, MARSHAL agents trained from Qwen3-4B develop strong strategic abilities, with up to 28.7% performance improvements in held-out games. More importantly, the capability acquired through self-play generalizes beyond games, yielding consistent performance gains of MASs in reasoning benchmarks. When integrated into leading MASs, our MARSHAL agent achieves significant zero-shot performance gains of up to 10.0% on AIME, 7.6% on GPQA-Diamond, and 3.5% on average across all benchmarks. These results establish self-play in strategic games as a powerful approach for developing generalizable multi-agent reasoning capabilities in LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2510_15414
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs
Yuan, Huining
Xu, Zelai
Tan, Zheyue
Yi, Xiangmin
Guang, Mo
Long, Kaiwen
Hui, Haojia
Li, Boxun
Chen, Xinlei
Zhao, Bo
Zhang, Xiao-Ping
Yu, Chao
Wang, Yu
Artificial Intelligence
Developing Large Language Models (LLMs) to cooperate and compete effectively within multi-agent systems (MASs) is a critical step towards more advanced intelligence. While reinforcement learning (RL) has proven effective for enhancing reasoning in single-agent tasks, its extension to multi-turn, multi-agent scenarios remains underexplored due to the challenges of long-horizon credit assignment and agent-specific advantage estimation. To address these challenges, we introduce MARSHAL, an end-to-end RL framework that incentivizes Multi-Agent Reasoning through Self-play witH strAtegic LLMs in both cooperative and competitive games. MARSHAL features a turn-level advantage estimator that aligns learning signals with each interaction for credit assignment, and an agent-specific advantage normalization to stabilize multi-agent training. By learning with self-play across cooperative and competitive games, MARSHAL agents trained from Qwen3-4B develop strong strategic abilities, with up to 28.7% performance improvements in held-out games. More importantly, the capability acquired through self-play generalizes beyond games, yielding consistent performance gains of MASs in reasoning benchmarks. When integrated into leading MASs, our MARSHAL agent achieves significant zero-shot performance gains of up to 10.0% on AIME, 7.6% on GPQA-Diamond, and 3.5% on average across all benchmarks. These results establish self-play in strategic games as a powerful approach for developing generalizable multi-agent reasoning capabilities in LLMs.
title MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs
topic Artificial Intelligence
url https://arxiv.org/abs/2510.15414