great suggestion
Technology:
questions below and will release the trained model together with code as suggested by reviewers to contribute to the community
We will release our trained model together with code as suggested.
Response to Review # 2: 2 On the downstream tasks: It seems like there is a misunderstanding on the downstream tasks we considered in this
We thank all reviewers for their thorough reading and valuable comments! Please find below our responses. In some datasets, the performance gap is even huge (e.g., more than 3% absolute On data splits: we did mention the data split statistics in Line 242-244. We will move the necessary data statistics to the main text in the revision. In addition, we provided the source code and instructions as supplementary files for better reproducibility.
To all reviewers We would like to thank the reviewers for their thoughtful comments and useful suggestions, which
Specifically we will compare ours to a kernel SVM by Zafar et al. and include this result in a