Fetching the paper…

Spec-LLaVA: Accelerating Vision-Language Models with Dynamic Tree-Based Speculative Decoding · Around