ChEBI-20

该数据集包含33,010个分子与描述的配对样本，这些样本被分为训练集、验证集和测试集。该数据集被视为评估的黄金标准，用于强制模型关注描述的语义层面。规模上，共有33,010个配对样本，任务是对分子描述的微调与生成进行评估。

This dataset contains 33,010 paired samples of molecules and their corresponding descriptions, which are divided into training, validation, and test sets. It is regarded as a gold standard for evaluation, designed to force models to focus on the semantic level of the descriptions. In terms of scale, there are a total of 33,010 paired samples, and its core task is to evaluate the fine-tuning and generation of molecular descriptions.

搜集汇总

背景与挑战

背景概述

该数据集是ChEBI-20化学分子数据集，包含分子描述、分子图结构、子结构标记和嵌入向量等多种化学信息表示形式，适用于分子表示学习和化学信息处理任务。