Baseline implementation questions

[Yang Wang]

Hi，authors

I've read the paper. There are some questions really puzzle me.
According to the description in the paper, the baseline was trained on several large datasets, and the testing was conducted on datasets that had never been trained on. The dimensions of the ARC data have been aligned, so it can be directly tested on the test dataset.

However, for the baseline models, different datasets are not the same, and the parameters vary from different datasets. Plus, the data of other baselines hasn't been aligned. So, the baseline method probably can't predict directly on the test dataset, right?

Therefore, I'm wondering what the experimental setup of the baseline was like in the article you published. (Was the aligned data in this repo used directly?)