We couldn't help but notice that one group has submitted near-perfect scores already. Are those results realistic and should we take them seriously?
Kind regardsPosted by: miwytt @ March 31, 2023, 2:02 p.m.
Dear participant. If I understood well, those results are from the Practise stage, in which the labels for both training and testing are available (you can check the files here: https://codalab.lisn.upsaclay.fr/competitions/10173#participate). Therefore, I suppose that this user has uploaded results from a over-fitted model (or directly, a file with the grounding truth).
The test labels from the next stage (Evaluation) will not made public until the competition end.
Please, do not hesitate in contact us if any other concernPosted by: JoseAGD @ March 31, 2023, 2:16 p.m.
That makes sense - thanks for your quick reply!Posted by: miwytt @ March 31, 2023, 2:22 p.m.