Benchmark RadarAI BENCHMARK PROFILE
Lost in Speech
A multilingual spoken hallucination detection benchmark with 12,013 news samples across English, Russian, and Kazakh, including synthetic and real-world fake news items in text and audio.
- Released
- 2026-08-25
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the gap in spoken hallucination detection, especially for low-resource languages, and evaluates transfer from synthetic to real-world fake news.
Motivation
While text-based hallucination detection has been extensively studied, spoken hallucination detection remains largely unexplored, particularly for low-resource languages.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.