Dental Appointment Scheduling Multi-Turn Speech Benchmark

A 25-turn multi-turn speech-to-speech benchmark for evaluating voice AI models as a dental office receptionist. It is part of the Audio Arena suite of 6 benchmarks spanning 221 turns across different domains. The benchmark was built by Arcada Labs.

Use Cases

• Evaluate multi-turn conversational performance of voice AI models on a dental office receptionist task.
• Assess speech-to-speech model capability in handling confusable patient names like Daniel and Danielle Nolan.
• Benchmark tool use and dialogue state tracking within a 25-turn appointment scheduling scenario.

Strengths

• Provides a focused 25-turn conversational evaluation scenario.
• Part of a larger suite of 6 benchmarks totaling 221 turns across domains.
• Specifically designed to test confusable name handling, a known challenge for voice AI.

Limitations

• Limited to a single, specific domain of dental appointment scheduling.
• The benchmark size is confined to 25 turns, which may not reflect longer, more complex real-world interactions.
• Geographic and linguistic scope is focused on US English, limiting generalizability.

Provenance

Source

Arcada Labs via Hugging Face.

Collection Method

Constructed as a benchmark for evaluation; specific data gathering method is unknown.

Freshness

Last updated on 2026-03-24.

Geography

United States (based on 'Regionus' tag).