Runs against a curated set of datasets, not arbitrary uploads — every result on this page is reproducible and every dataset below is chosen to show something specific.
Time-series data isn't supported. Model evaluation assumes rows are independent — if row order matters (stock prices, sensor logs), results here would be misleading.
Read the dashboard flags. Every result shows a baseline comparison and held-out test methodology next to the scores — a high number next to a weak baseline isn't the same as a strong model.