Independently confirmed proposing a reshuffle + rescore.
I can reproduce this finding. The practical effect is that genuine modelling signal is masked on the public leaderboard. Could the organizers please shuffle the test row order and rescore the existing submissions? That removes the exploit without altering anyone's actual predictions, and restores the leaderboard as a measure of real model quality
haha. always nice when things are like
🤣🤣
😂
Independently confirmed proposing a reshuffle + rescore.
I can reproduce this finding. The practical effect is that genuine modelling signal is masked on the public leaderboard. Could the organizers please shuffle the test row order and rescore the existing submissions? That removes the exploit without altering anyone's actual predictions, and restores the leaderboard as a measure of real model quality
Thanks for catching and raising this.
¯\_(ツ)_/¯
😂😂
ABSOLUTE CINEMA!! 😂😂