IVRA: Improving Visual-Token Relations for Robot Action Policy with Training-Free Hint-Based Guidance

返回详情 VLA / Vision-Language-Action 每日论文卡