I would like to add that, in fact, uploading a .json file also leads to an error. It seems the online evaluation system might not be set up yet.
Additionally, regarding the constraint mentioned in track I, there seems to be a discrepancy. The guidelines note a hard constraint of 100 tokens for each jailbreak prompt, whereas the utils.py checks for the number of added tokens exceeding 100 (if num_added_token > 100). Could the organizers please clarify which guideline we should follow?
Appreciate your efforts and looking forward to your guidance.
Best,
Chao