Could you confirm that the columns we saw in the sample test data also appear in the real test set? For instance, we suspect sponsor columns such as intsponsorid_p2_intclinicaltrialids or DrugKey are not in the real test set. We hope it’s our own bug, but we have a strong suspicion the test set might be problematic. Thanks!
Hi, you are correct. The dataset available during evaluation was one version older and didn’t contain columns like intsponsorid_p2_intclinicaltrialids
/DrugKey
.
This has been fixed now and current success submissions’ score have been recalculated.
thanks for the quick work there shivam.