Reproduce “InteractionBench: Measuring Timing, Task Accuracy, Whisper and Volume, and Grounding in Real-Time Voice Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Richard Yucheng He, Chen Xu, Yihang Liu, Tairan Chen, Baodong Cao
InteractionBench evaluates how real-time voice assistants handle conversational timing, spoken-task answers, voice output, and groundedness across long exchanges. It replays shared recorded user audio to systems, records both sides on one timeline, and scores four dimensions separately. In this evaluation, a turn-based cascade passed 38/50 spoken-task items (76%) but only 68/140 timing items (49%); results are specific to this setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.