TY  - RPRT
TI  - KVBoost: Chunk-Level Key-Value Cache Reuse with Deviation-Guided Recomputation for Efficient Large Language Model Inference
AU  - Srihari Unnikrishnan
PY  - 2026
UR  - https://arxiv.org/abs/2608.21362
ID  - 2608.21362
ER  -