MultiTalkBench: Bilingual Multi-Party Dialogue Benchmark

MultiTalkBench is a benchmark dataset for evaluating speech-to-speech dialogue systems. It tests systems on long interactions, multi-party conversations, and Chinese-English bilingual ability. The dataset was created by MultiTalk and last updated on 2026-05-06.

Use Cases

• Evaluate long-range entity tracking in conversations based on explicit probes for long interactions.
• Benchmark turn-taking and addressee-selection performance based on multi-party interaction metrics.
• Test bilingual speech-to-speech system fluency based on Chinese-English dialogue ability.
• Assess topic coherence over extended conversations based on probes for long interactions.

Strengths

• First benchmark to jointly evaluate long, multi-party, and bilingual full-duplex dialogue.
• Includes explicit probes for long-range entity tracking and topic coherence.
• Provides quantitative metrics for addressee-selection and turn-taking.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is 2026-05-06.

Provenance

Source

MultiTalk

Freshness

Last updated 2026-05-06 10:23:21.