Question about the training dataset

[ZinYY]

Hi, thanks for the great work!

I noticed that in the provided training dataset (data/train_data.json), some entries have identical query and response fields (for example, the item starting with “How many moles of hydrochloric acid…”). Could you please clarify why? Thanks a lot!

[JinaLeejnl]

@ZinYY data/train_data.json only provides sample data. The full training dataset is available here: https://huggingface.co/datasets/GSAI-ML/ReFusion. All our data comes directly from high-quality open-source datasets without any post-processing, except for length limits.