KMMAU: Korean Multi-Modal Audio Understanding Benchmark

KMMAU is a Korean Multi-Modal Audio Understanding benchmark designed to evaluate speech models. It was created by KRAFTON and covers 9 subsets of tasks including age and gender identification, fact extraction, and topic summarization. The dataset page was last updated on 2026-06-01.

Use Cases

• Benchmarking model performance on Korean audio comprehension based on described task subsets
• Evaluating speech model robustness across different speaker demographics like age and gender
• Testing multimodal AI capabilities for tasks like fact extraction and topic summarization from audio

Strengths

• Covers 9 distinct subsets of audio understanding tasks
• Designed specifically for evaluating Korean speech-language models
• Created by the organization KRAFTON

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

KRAFTON

Freshness

Last updated 2026-06-01 05:14:35; freshness should be verified

Geography

Korean language focus