Submission Rule Adjustment to Ensure Fairness Across Both Tasks

Submission Rule Adjustment to Ensure Fairness Across Both Tasks  

  By: Kobalt on Sept. 4, 2024, 11:25 a.m.

Dear autoPET III Organizers,

We are writing as a participating team in the autoPET III challenge, and we would like to raise your attention to a potential issue regarding the current submission rules that may unintentionally disadvantage teams aiming to address both the model-centric and data-centric tasks.

As it currently stands, each team is allowed to submit only two models, which can be distributed either as 1-1 across both categories or both models to a single category. This rule places teams that wish to contribute to both the model-centric and data-centric tasks at a disadvantage. Specifically, teams focusing on only one category can submit their two best models for that category, maximizing their chances for success. In contrast, teams like ours, which are contributing to both tasks, must divide our submissions, potentially decreasing our chances in either category.

To ensure a more balanced competition and to encourage comprehensive participation across both tasks, we propose to limit the number of submissions for each team to either 1 or 2 per category instead of a total submission limit. This adjustment would enable teams to compete fairly in both tasks without being penalized for addressing multiple aspects of the challenge. We believe this change would increase the number of final submissions for both tasks without strategic disadvantages.

We hope you will consider this suggestion and look forward to your response.

Best regards,

DKFZ Team

Re: Submission Rule Adjustment to Ensure Fairness Across Both Tasks  

  By: michael.ingrisch on Sept. 6, 2024, 11:30 a.m.

Dear DKFZ team,

Thank you for your feedback regarding the submission rules. We understand your concern about the current model submission limitations, and we agree that this is a valuable point for consideration. In future challenges, we will certainly take your suggestion into account to ensure a more balanced competition for teams addressing both model-centric and data-centric tasks. Additionally, please note that models submitted in the data centric category are also eligible for ranking in category 1, which may help mitigate the concerns you raised.

However, given the proximity of the final submission deadline next week, we believe that making any rule changes at this stage would be unfair to other participating teams who have been preparing under the current guidelines. We appreciate your understanding and look forward to receiving your submissions.

Best regards,

the organizing team