Audio Evaluation Datasets for Speech and Understanding Tasks

A 2025 collection of multiple audio datasets compiled by XiaomiMiMo for the MiMo-Audio-Eval toolkit. It includes datasets for automatic speech recognition, text-to-speech, and audio understanding tasks such as AISHELL1, LibriSpeech, and SeedTTS.

Use Cases

• Benchmark automatic speech recognition models using the AISHELL1 and LibriSpeech datasets.
• Evaluate text-to-speech synthesis systems with the SeedTTS dataset.
• Test audio understanding and reasoning models on the MMAU and MMSU datasets.

Strengths

• Compiled by XiaomiMiMo, a major technology organization.
• Includes established benchmark datasets like AISHELL1 and LibriSpeech.
• Updated in September 2025, indicating recent maintenance.

Limitations

• Specific dataset sizes, row counts, and column structures are unknown.
• The composition and balance of the included datasets are not detailed.
• File formats and data accessibility details are unspecified.

Provenance

Source

XiaomiMiMo

Collection Method

Collection of existing audio datasets for an evaluation toolkit.

Freshness

Last updated on 2025-09-18.