Yi Ding

RL checkpoints of Octopus-8B and baselines of paper: Learning Self-Correction in Vision–Language Models via Rollout Augmentation