Reproduce “Voice-Native LLMs Hear Frustration but Do Not Listen to It” 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
Hanjun Luo, Yuhan Lu, Hanan Salam
The authors test whether voice assistants perceive mild vocal frustration separately from wording, and whether it affects answers. Across eight models, neutral-wording clips with frustrated voices are labeled frustrated in at most 3.1% of cases, while seven detect strong acted anger. For two open-weight models, probes decode vocal condition from audio states; findings concern mild synthesized English frustration.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.