作者位次 #7/10ACL 2026Poster· Natural Language Processing
MARS ^2 : Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation
Pengfei Li · Shijie Wang · Fangyuan Li · Yikun Fu · Kaifeng Liu · Kaiyan Zhang · Dazhi Zhang · Yuqiang Li · Biqing Qi · Bowen Zhou
摘要
Reinforcement learning (RL) paradigms have demonstrated strong performance on reasoning-intensive tasks such as code generation. However, limited trajectory diversity often leads to diminishing returns, which constrains the achievable performance ceiling. Search-enhanced RL alleviates this issue...