Dear all, I wanted to inform you that there are some errors in the mixture definition of the leaderboard set (as well as the training dataset) Most of the mistakes appear mainly in the Bushdid dataset. Here are a few examples: 1- In Bushdid Mixture Label 7 the CID==12473 has the CAS# of 634-97-9 (https://pubchem.ncbi.nlm.nih.gov/compound/Pyrrole-2-carboxylic-acid) This molecule is not present in the compound used in Bushdi experiment (https://www.science.org/doi/full/10.1126/science.1249168). If you dig deeper into the mixture definition in the original paper you end up with gamma-valerolactone with CID== 7921 and CAS#==108-29-2. Thus 12473 should be replaced by 7921 2-Some molecules are missing in the definition of mixtures. For example, in Bushdid Mixture Label 203 the CID of the second molecule is labeled 0, which again if you look into the original dataset it should be CID==1127 Here are a few more replacement examples for CID in the leaderboard set: Original ----> Correction 22311 ---> 440917 14896 ---> 440967 25137858 ---> 91497