KV-Efficient VLA: A Method to Speed up Vision Language Models with RNN-Gated Chunked KV Cache

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