SciKnowEval
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/ADaM-BJTU/OpenRFT
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个新近发布的科学基准，涵盖了五种不同的能力，特别强调知识推理水平（L3）。它包含了多个子集，包括GB1-ftness预测、逆合成、化学计算、分子结构预测、高中物理计算、材料计算、扩散速率分析以及钙钛矿稳定性预测。这些子集涵盖了生物学、化学、物理和材料科学四个领域，任务旨在跨多个领域进行知识推理。
This dataset is a newly released scientific benchmark covering five distinct capabilities, with a particular emphasis on knowledge reasoning level (L3). It includes multiple subsets, such as GB1-fitness prediction, retrosynthesis, chemical calculation, molecular structure prediction, high school physics calculation, materials calculation, diffusion rate analysis, and perovskite stability prediction. These subsets span four disciplines: biology, chemistry, physics, and materials science, and the tasks are designed to conduct knowledge reasoning across multiple domains.
提供机构：
Not explicitly stated
