Turnbench: A Multi-Domain Benchmark for Turn-Taking Dynamics in Spoken Dialogue

Back to Devoured - September 15, 2026

DEVOURED

Turnbench: A Multi-Domain Benchmark for Turn-Taking Dynamics in Spoken Dialogue

AI research Arxiv

TURNBENCH is a massive 134-hour benchmark set designed to evaluate how AI agents handle the complexity of spoken dialogue, specifically turn-taking and interruptions.

What: The dataset includes a 30-hour benchmark for detection accuracy and 104 hours of training data across six domains, aimed at improving real-time voice interactions.

Original article

TURNBENCH introduces a 30-hour, six-domain benchmark for end-of-turn and interruption detection, plus a 104-hour training set.

Explore all 47 articles in today's digest