Fetching the paper…

VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching · Around