Christopher Lott

QuoKA: Query-Oriented KV Selection for Efficient LLM Prefill

Training-Free Multi-Token Prediction via Probing

KeyDiff: Key Similarity-Based KV Cache Eviction for Long-Context LLM Inference in Resource-Constrained Environments

How efficient is LLM-generated code? A rigorous & high-standard benchmark

CAOTE: Efficient Caching through Attention Output based Token Eviction