作者位次 #8/12ACL 2026Poster· Natural Language Processing
I ² B - LPO : Latent Policy Optimization via Iterative Information Bottleneck
Huilin Deng · Hongchen Luo · Yue Zhu · Long Li · Zhuoyue Chen · Xinghao Zhao · Ming LI · Chuyang Zhao · Jihai Zhang · MengChang Wang · Yang Cao · Yu Kang
摘要
Despite recent advances in Reinforcement learning with verifiable rewards (RLVR) for large language model (LLM) reasoning, most methods suffer from exploration collapse, as the semantic homogeneity of random rollouts traps models in narrow, over-optimized behaviors. Existing methods leverage poli...