BioDSA-1K

该数据集名为BioDSA-1K，是一个包含1,029项以假设为中心的任务和1,177项分析计划的基准测试集，这些内容是从超过300项已发表的生物医学研究中精心挑选出来的。它旨在评估人工智能代理在真实生物医学假设验证任务上的表现。BioDSA-1K数据集不仅包含不可验证的假设，还涵盖了多种研究类型，如基因组学、综合性研究、治疗学、生物标志物、转化研究和分子研究。该数据集规模宏大，包含来自超过300项研究的1,029项任务和1,177项分析计划，其任务重点在于假设验证和数据分析。

The dataset named BioDSA-1K is a benchmark collection comprising 1,029 hypothesis-centric tasks and 1,177 analysis protocols, which are meticulously curated from over 300 published biomedical studies. Its core purpose is to evaluate the performance of AI Agents on real-world biomedical hypothesis verification tasks. BioDSA-1K not only includes unverifiable hypotheses, but also covers a variety of research types such as genomics, comprehensive studies, therapeutics, biomarkers, translational research and molecular research. Boasting a substantial scale, the dataset encompasses 1,029 tasks and 1,177 analysis protocols originating from more than 300 studies, with the tasks primarily focusing on hypothesis verification and data analysis.