KVGuard: A KV Cache Safety Checker for LLM Serving Systems

Researchers at Seoul National University present KVGuard, a debugging-time checker for the KV cache in LLM serving. It keeps a small shadow record per PagedAttention block, tracking which requests may access it and which entries were written, and runs five checks from CPU allocation through GPU access. The prototype on vLLM and SGLang covers 29 historical bugs and found 8 new ones, with no false positives reported. Overhead is 3.0% and 7.4% TPOT on one model, Qwen 2.5 14B. These figures are the authors' own and are not independently verified.