Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation cs.CV · 2026-05-18