Evaluation Results

Each team was allowed to submit multiple entries. The public leaderboard reflected performance on 70% of the test set, while the remaining 30% was kept private by the organizers. The values presented in this table were recalculated based on each team's highest-scoring entry on the private leaderboard, using the entire test set (both the public and private portions). Therefore, these recalculated results may differ slightly from the original public leaderboard results. Additionally, since some teams only participated in predicting specific columns, scores for columns without submissions are marked as ``--.''