-
R-lens: Making J-lens More Faithful on Early Layers
R-Lens
-
Verbalizable Representations Form a Global Workspace in Language Models
J-Lens and J-Space
-
Scaling RL training in the age of multi-agent systems
Raymond Feng AC2
-
InstructPix2Pix: Learning to Follow Image Editing Instructions
InstructPix2Pix
-
Video Generation Models are General-Purpose Vision Learners
GenCeption