Benchmark RadarAI BENCHMARK PROFILE
StanceBench
StanceBench evaluates interpersonal stance in conversational speech across 9 dimensions using LLM-as-a-judge on the Seamless Interaction corpus.
- Released
- 2026-06-27
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the gap in benchmarks for prosody and interactional nuance in speech-to-speech models, focusing on judge bias and robustness.
Motivation
Speech-to-speech dialogue models increasingly depend on prosody and interactional nuance to convey social intent, yet benchmarks for these cues remain limited.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.