Fetching the paper…

D2O: Dynamic Discriminative Operations for Efficient Long-Context Inference of Large Language Models · Around