Private Leaderboard

A Private Leaderboard is a secure ranking system used in competitions or machine learning challenges, where participants submit their models or solutions. It reveals the performance of these submissions on a hidden or reserved data set that the participants haven't seen during development. This approach helps ensure the evaluation is unbiased and prevents overfitting—when a model performs well only on known data. The Private Leaderboard typically remains hidden during the challenge and is revealed at the end, providing an accurate assessment of which models truly generalize well to new, unseen data.