sumuks/yourbench_y1

YourBench Y1是一个精心策划的数据集，包含来自8个不同领域的文档，专门设计用于评估语言模型在2024年7月后生成的内容上的表现。该数据集为测试模型在当代内容上的表现提供了独特的基准，涵盖了多样化的专业和技术领域。每个领域包含5个文档，总共40个文档。每个文档包括完整内容和GPT-4生成的摘要。

YourBench Y1 is a carefully curated dataset of documents from 8 different domains, specifically designed to evaluate language models on content likely generated or produced after July 2024. This dataset provides a unique benchmark for testing model performance on contemporary content across diverse professional and technical domains. The dataset includes 40 documents, each with full content and a GPT-4-0824 generated summary. The average content length is approximately 10,756 tokens, and the summary length is about 130 tokens. The dataset covers 8 domains including corporate, financial, government, health, legal, misc, news, and research.