VLM4VLA: REVISITING VISION-LA | Pangram Labs