The submission files will be in CSV format. The files must include the original idx, context_idx, and label fields. The idx and context_idx must follow the original order from dev and eval datasets.
See the example from the submission file on the right: