Multi-reward reinforcement learning offers a promising framework for aligning large language models (LLMs) with diverse objectives, but conflicting reward si…
机构：中山大学
来源：arXiv 2610.35480 | AI4Papers 论文推荐平台