Sergey Levine (@svlevine)
2025-05-28 | โค๏ธ 466 | ๐ 63
Fun project at PI: knowledge insulation for VLAs. We figured out how to train VLAs with cont. actions much more effectively by insulating the VLM and training it with discrete actions, while action expert learns on top. 5-7x faster, and importantly way better language following ๐
๋ฏธ๋์ด
