04:00
2026-09-04
arxiv.org
artificial-intelligence
DuplexSpeechBench-IFEval: Evaluating Implicit Instruction Following in Full-Duplex Voice Agents
A new benchmark, DuplexSpeechBench-IFEval (DSB-IFEval), evaluates implicit instruction-following in full-duplex voice agents, comprising 1,038 test cases across eight assistant roles. Testing six realβ¦