Dalton Jones

QuoKA: Query-Oriented KV Selection for Efficient LLM Prefill

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

PADRe: A Unifying Polynomial Attention Drop-in Replacement for Efficient Vision Transformer

CAOTE: Efficient Caching through Attention Output based Token Eviction