2026
Where Signals Are Sparse, We Synthesize: Reinforcing Self-Corrective Reasoning in Vision–Language Models via Rollout Augmentation
ICML 2026poster
Self-correction is essential for solving complex reasoning problems in vision–language models (VLMs), yet existing reinforcement learning (RL) methods struggle to learn it. Effective self-correction behaviors emerge only rarely during RL, making learning signals sparse. To address this challenge, we…