Why Does RL Generalize Better Than SFT? A Data-Centric Perspective on VLM Post-Training

Add code
Feb 11, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: