Dear Organizing Committee,
We have noticed that this year’s competition dataset partially overlaps with last year’s validation set, which is still accessible through historical links. Given this, we have two questions:
Are participants allowed to use the validation set for training in this year’s KVQ competition?
During the evaluation phase, how can we effectively determine whether a participant’s submission has been trained on the validation set, leading to potential label leakage?
We appreciate your clarification. Thank you!
Thanks for your questions. Yes, you can utilize the validation for training in the test stage, I will release its label together.
Posted by: Xin.Li @ March 16, 2025, 1:35 p.m.