FAST: Efficient Action Tokenization for Vision-Language-Action Models (Paper Walkthrough)
Download (MP3)
Bagikan
Facebook
Twitter