**Scoring and test dataset details**: For this track, 20% of the Sygen data will be held back as test data and used in the evaluation process. The Kaggle public leaderboard is computed on a subset of the test data for the functional prediction task, so while it will give you a sense of your relative performance, it may not exactly match the final evaluation. This prevents overfitting through submissions. Also note that the leaderboard is based on a simple root mean squared error across all entries, including assessment above the NLI. 

**Imputation performance**: For imputation of the ISNCSCI scores, for each test-set sample, the final prediction will be scored by first measuring the error between the predicted and true scores. Then, the sum of those residuals will be computed to provide a score for each individual. This will be averaged across individuals in the test data set to determine the final score. In short, we will base evaluations on a root mean squared error between true and predicted scores.

We provide a template for the submission file in the Data section.