VoiceBench
收藏资源简介:
VoiceBench是由新加坡国立大学开发的第一个用于多方面评估基于大型语言模型(LLM)的语音助手的基准数据集。该数据集包含1817条数据,涵盖了真实和合成的语音指令,旨在评估语音助手在不同实际场景中的表现。数据集的创建过程包括利用先进的文本到语音(TTS)和语音克隆模型生成多样化的语音样本,并模拟不同的环境效果。VoiceBench主要应用于语音助手的性能评估,旨在解决当前评估方法在复杂、真实世界场景中的不足,提供一个全面的评估框架。
VoiceBench is the first benchmark dataset developed by the National University of Singapore for multi-faceted evaluation of large language model (LLM)-based voice assistants. It contains 1,817 data entries, covering both real and synthetic voice commands, and aims to evaluate the performance of voice assistants across various real-world scenarios. The dataset was created by leveraging state-of-the-art text-to-speech (TTS) and voice cloning models to generate diverse speech samples, as well as simulating different environmental effects. Primarily applied to the performance evaluation of voice assistants, VoiceBench is designed to address the limitations of current evaluation methods in complex real-world scenarios, providing a comprehensive evaluation framework.
VoiceBench 数据集概述
数据集描述
VoiceBench 数据集用于评估基于大型语言模型(LLM)的语音助手。该数据集托管在 Hugging Face 上,可通过 Hugging Face 的 datasets 库进行访问和集成。
数据集访问
数据集的访问链接为:VoiceBench Dataset。
数据集使用方法
在 Python 环境中加载数据集的示例如下:
python from datasets import load_dataset
加载 VoiceBench 数据集
可用的子集:alpacaeval, commoneval, sd-qa, ifeval, advbench
dataset = load_dataset("hlt-lab/voicebench", alpacaeval)
引用信息
如果在研究中使用 VoiceBench 数据集,请引用以下论文:
@article{chen2024voicebench, title={VoiceBench: Benchmarking LLM-Based Voice Assistants}, author={Chen, Yiming and Yue, Xianghu and Zhang, Chen and Gao, Xiaoxue and Tan, Robby T. and Li, Haizhou}, journal={arXiv preprint arXiv:2410.17196}, year={2024} }

- 1VoiceBench: Benchmarking LLM-Based Voice Assistants新加坡国立大学 · 2024年



