ResearchBenchmarksConversational voice agent evaluation on User interaction study Direct, RAG, MCPFollow3.44LatencyBase SLM3.4083.6243.844.056Nov 10, 2025Evaluation ResultsMethodMethodLinksLatencyClarityFluencyResponse LengthCoherenceTask CompletionNaturalnessSatisfactionBase SLM2025.113.444.533.6932.942.333.612.03Frontier2025.113.464.814.834.074.724.354.413.8ConvFill2025.114.244.694.563.944.284.313.833.81