ResearchBenchmarksSemantic Caching on QNLIFollow1,504.43End-to-End Latency (min)MVR-cache1,458.65521,767.63512,076.6152,385.5949May 24, 2026Evaluation ResultsMethodMethodLinksEnd-to-End Latency (min)Algorithm Time (min)MVR-cache2026.051,504.4317.62vCache2026.051,53614.1ColBert2026.051,626.3739.28POQD2026.052,648.81,048.48