Fetching the paper…

LOOK-M: Look-Once Optimization in KV Cache for Efficient Multimodal Long-Context Inference · Around