Reproducing the CEDAR and IEDB benchmarks: random seeds and data splits

[Yongrui-Wang]

We are attempting to reproduce the CEDAR and IEDB benchmark experiments reported in the ImmunoStruct paper.

The manuscript states that the experiments were performed under five random seeds. We could not identify the exact five seeds or the corresponding train/validation/test split indices for either the CEDAR or IEDB experiments in the repository.

Could you please provide either:

1. the exact five random seeds used for the reported CEDAR and IEDB experiments, or preferably
2. the corresponding train/validation/test indices (or split files) for each of the five runs?

Having the exact splits would help us reproduce the reported benchmark results under the same evaluation setting.

Thank you very much for making the code, datasets, and model weights publicly available.

[ChenLiu-1996]

We used seeds 1,2,3,4,5 for the experiments. We have rerun the experiment on a different device and the results seem similar to what was reported in the paper.