Jasoncsc/DCA-Bench
收藏Hugging Face2024-06-12 更新2024-06-12 收录
下载链接：
https://hf-mirror.com/datasets/Jasoncsc/DCA-Bench
下载链接
链接失效反馈官方服务：
资源简介：
DCA-Benchmark旨在为评估LLM代理在在线数据集平台上发现数据质量问题的能力提供一个全面的基准。该数据集包含来自8个在线数据集平台的91个代表性样本，并根据内容和难度分为4种类型和18个标签。关键特点包括真实案例、多难度级别和自动评估方案。
The DCA-Benchmark dataset aims to provide a comprehensive benchmark for evaluating the capabilities of large language model (LLM) agents in discovering data quality issues across online dataset platforms, representing the first step of the curation pipeline. The dataset includes 91 representative samples collected from 8 online dataset platforms, classified into 4 types with 18 tags according to their various content and difficulty. The dataset features real-world cases with minimal simplification, multiple difficulty levels, and an accurate automatic evaluation scheme using GPT-4 to replace human annotators.
提供机构：
Jasoncsc
原始信息汇总
数据集许可证信息
- 许可证类型: Apache-2.0
