Survey of System-Aware KV Cache Optimization for LLM Serving
July 10, 2026
This survey categorizes KV cache optimization strategies into temporal, spatial, and structural dimensions to improve LLM inference efficiency. It analyzes how execution, placement, and representation co-design impacts memory intensity and latency in serving infrastructures.
HOW THIS AFFECTS YOU
●
builderUse these structural and temporal optimization patterns to reduce inference costs and latency.