M3-DuplexBench evaluates multilingual full-duplex spoken dialogue with turn-taking, backchannels, and user interruptions.
unassessed
| Category | generation |
|---|---|
| Subcategory | spoken dialogue |
| Metric | dialogue performance |
| Direction | higher_is_better |
| Unit | score |
| Publisher | M3-DuplexBench authors |
M3-DuplexBench tests spoken dialogue systems that listen while speaking. It covers English and Japanese, casual conversation and multi-turn question answering, and multiple context conditions for studying dialogue history.
Multi-turn audio dialogue with full-duplex interaction and context variants.
No model card in ModelSpec reports this benchmark yet.