Skip to content

Dataset · tencent

Real-world speech models benchmark

tencent/WildSpeech-Bench

Tencent's first benchmark for end-to-end speech-to-speech capabilities of speech LLMs.

What's inside

The dataset contains between 1 and 10 thousand audio and text examples in English. It is designed for comprehensive evaluation of SpeechLLM class models in speech-to-speech mode under near-real-world conditions. The repository also provides evaluation code, but details on file formats and recording sources are not specified in the snippet.

What you can do with it

  • Testing and comparing end-to-end speech-to-speech models
  • Checking speech recognition and generation quality under complex conditions
  • Validating multimodal voice assistants
Languages
English
Size
1K–10K rows
Downloads this month
358
Likes
37

How to load it

from datasets import load_dataset

ds = load_dataset("tencent/WildSpeech-Bench")