Fetching the paper…

LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models · Around