Token-level Response-visual Attention Guidance for Multimodal LLMs Knowledge Distillation got accepted to ECCV 2026.