KVzip: Query-Agnostic KV Cache Compression with Context Reconstruction

NeurIPSOral2025

Authors
Jang-Hyun Kim, Jinuk Kim, Sangwoo Kwon, Jae W. Lee, Sangdoo Yun, Hyun Oh Song
Affiliation
Apple
Venue
NeurIPS 2025
Track
Oral

TL;DR

We propose a novel query-agnostic KV cache eviction method for multi-query scenario.

Opening excerpt from the authors’ abstract. source

Read the paper

← All NeurIPS 2025 Oral papers · Browse the whole archive