Training-Free Acceleration for Vision-Language-Action Models with Action Caching and Refinement | Digital Library | PAMCET | PAMCET