Xianhang Li*

Latest

• What If We Recaption Billions of Web Images with LLaMA-3?
• Revisiting Adversarial Training at Scale
• An Inverse Scaling Law for CLIP Training