Feb 1 • 💬 General
DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning
*** DINO-World Model ***
New work from @gaoyuezhou, @gary_phkkk, @lerrelpinto and @yannlecun.
A system capable of planning complex action sequences using an action-conditioned world model trained from off-line data.
Pure planning (no reinforcement learning whatsoever).
It uses a JEPA architecture in which the image encoder is DINOv2, and the action-conditioned predictor trained from off-line data.
Website: https://dino-wm.github.io/
2
1 comment
DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning
skool.com/data-alchemy
Your Community to Master the Fundamentals of Working with Data and AI — by Datalumina®
Leaderboard (30-day)
1
+203
2
+75
3
+33
4
+32
5
+28
Powered by