Mac Schwager (@MacSchwager)

2025-11-12 | โค๏ธ 73 | ๐Ÿ” 13


How do we fuse VLM physical intuition with runtime adaptation in robot manipulation? It requires careful uncertainty quantification to properly weight the two. This is the key behind @maggi3wang โ€˜s Phys2Real approach. Congrats Maggie and team on this insightful work!


์ธ์šฉ ํŠธ์œ—

Maggie Wang (@maggi3wang)

Humans estimate object physical properties through vision and refine them through touch. ๐˜Š๐˜ข๐˜ฏ ๐˜ณ๐˜ฐ๐˜ฃ๐˜ฐ๐˜ต๐˜ด ๐˜ฅ๐˜ฐ ๐˜ต๐˜ฉ๐˜ฆ ๐˜ด๐˜ข๐˜ฎ๐˜ฆ?

Excited to share ๐—ฃ๐—ต๐˜†๐˜€2๐—ฅ๐—ฒ๐—ฎ๐—น ๐Ÿค–๐Ÿ‘๏ธโœ‹: VLM priors + interactive adaptation for uncertainty-aware sim-to-real! https://t.co/pmYpW1GOCw ๐Ÿงต (1/N) https://t.co/OgAUyo9csN

์›๋ณธ ํŠธ์œ—

๐ŸŽฌ ์˜์ƒ

Tags

Robotics AI-ML VLM