Shibattic
Currently in stealth
Physical AI · Tokyo
Research
2025
The Compression Gap: Why Discrete Tokenization Limits Vision-Language-Action Model Scaling
arXiv
↗
