Picture for Liuqun Zhai

Liuqun Zhai

SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference

Add code
Apr 23, 2026
Viaarxiv icon