Fetching the paper…

HiRE: High Recall Approximate Top-$k$ Estimation for Efficient LLM Inference · Around