New benchmark DualEvasion tackles vocal and textual evasion in earnings calls

Researchers have introduced DualEvasion, a new benchmark designed to detect evasion in earnings calls by analyzing both textual transcripts and vocal cues. This benchmark, comprising 505 annotated question-answer pairs from 60 earnings calls, labels textual evasion and speaker confidence. Experiments revealed that current multimodal models struggle to accurately identify vocal confidence, especially when responses are unconfident, and tend to interpret acoustic cues in isolation rather than relative to a speaker's baseline. AI

IMPACT This research could lead to more sophisticated AI models capable of detecting subtle forms of evasion in spoken communication.

RANK_REASON The cluster contains a research paper detailing a new benchmark for analyzing communication. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• arXiv
• DualEvasion

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →