all of the submissions from the same team are taken as the final result?

Dear Organizer,

Only the first successful submission of the participants will be considered as their final result. However, for example, if a team consists of 5 members, then they can submit 5 times to verify their model. Apparently, it is unfair to the individual participants.

So, I wanna ask, if all of the submissions from the same team are taken as the final result?

Best regards,

You said that:

However, after observing that participants encountered difficulties in the submission process, we have decided to allow up to three submissions per participant. However, it is important to note that only the first successful submission will be considered as the final result.

I wanna ask the first successful submission is the first successful submission of a participant or the first successful submission of a team?

If the anwser is a participant, then the rule is more beneficial to the team with many members. For example, if there are 5 members in the team, then they can submit [(5-1)*3 + 1 = 13] times.

During the testing phase, if multiple participants from the same group submit their entries, only the best score among the participants’ first submissions will be considered as the final result. As announced at the beginning of the challenge, the submission limit for both the training and testing phases is per participant. Furthermore, some participants have already submitted their algorithms based on the existing system.

Nevertheless, we will discuss the proposed concerns regarding the submission system with the organizers to ensure clarity in future challenges and carefully consider the suggested recommendations when constructing the guidelines for the next challenge. Thank you for your understanding.

Best regards,

Thanks for your reply.

But, again, if a team have 5 members. Then:

1# member submit and get the result using their model, then using the left 2 chances to verify the model and get the score;

2# member continue to improve and verify the 1#'s model;

3# improves and verifies the 2#'s model;

...

Then if the team have 5 members, they have 3*(5-1) + 1 chances to verify the model. Under this rule, the best way for a participant is to create 6 accounts and submit the model 16 times in the testing phase.

Then the testing phase may not reflect the effectiveness of one's model due to the probable over-fitting. So I suggest that only take the first submission of a team.

Best Regards,

I agree with this comment. In all the challenges I have participated, each group (either one or more participants) has only one submission opportunity.

Hey hello,

I also agree with this - individual participant here, so a bit biased against teams :)

The point is that probably additional submissions should only be allowed if the initial submission did not generate a score, otherwise you can start learning from that score and pass the knowledge to the next participant from your team to use it, as mentioned by elliotqii.

That said, I understand rules were set before and an a-posteriori discussion now makes little sense. As a possible fix, I propose that the organizers give the first prize considering the current system, and maybe an honorary prize to the winner when considering only one submission per team (the first valid submission per team, considering an inidividual as a team).

Cheers,

Adrian