Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models
返回详情
VLA / Vision-Language-Action 每日论文卡