[Bug] Score discrepancy and data contamination in Bioavailability_Ma leaderboard

[Taekya]

Describe the bug
There appears to be a score discrepancy and potential data contamination on the ADMET Leaderboard for the Bioavailability_Ma task. It seems that the submission values/scores for Pgp_Broccatelli were accidentally applied to Bioavailability_Ma, as the scores for several overlapping models across both leaderboards are identical.

To Reproduce
Steps to reproduce the behavior:

1. Go to the TDC ADMET Leaderboard.
2. Compare the model scores/rankings for the Bioavailability_Ma task.
3. Compare them with the scores for the Pgp_Broccatelli task.
4. See the identical scores across multiple overlapping models.

Expected behavior
The Bioavailability_Ma leaderboard should display the correct, independent evaluation scores for the submitted models based on the Bioavailability dataset, rather than duplicating the Pgp_Broccatelli results.

Screenshots

Environment:

• OS: N/A (Web platform issue)
• TDC version: Leaderboard Website

Additional context
First of all, thank you for maintaining the TDC Leaderboard; it has been incredibly useful for my research.

I noticed this issue while analyzing the benchmarks. Since the scores of several models perfectly match between the two completely different tasks (Bioavailability_Ma and Pgp_Broccatelli), it strongly suggests a data mapping or logging error during the leaderboard update.

Best regards,
Taekyeong An