Audio samples generated for the paper "Multi-Faceted Interactivity Alignment in Full-Duplex Speech Models". The samples were created using Full-Duplex-Bench v1 for static evaluation and v2 for real-time multi-turn dialogue, as described by Kyutai. The dataset was last updated on June 4, 2026.
Use Cases
- Evaluating speech model interactivity based on the Full-Duplex-Bench framework
- Comparing static versus real-time dialogue performance based on the described benchmark versions
- Auditory analysis of multi-turn alignment in conversational AI systems
- Benchmarking full-duplex speech model outputs against reference samples
Strengths
- Samples are directly linked to a published research paper and accompanying blog post
- Evaluation covers two distinct scenarios: static evaluation and real-time multi-turn dialogue
- Associated with specific model releases on the Hugging Face platform
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- Kyutai
- Collection Method
- Generated for evaluation benchmarks as described in the accompanying paper
- Freshness
- Last updated 2026-06-04 15:02:54; freshness should be verified