1
What Do People Actually Want From AI? Mapping Preference Plurality
顶会论文揭示RLHF聚合偏好的根本缺陷，系统绘制人类对AI的真实多元需求图谱
arXiv:2606.06674v1 Announce Type: new Abstract: Large Language Models (LLMs) are often fine-tuned through Reinforcement Learning from Human Feedback (…