tabpfn-classify

To establish a TabPFN classification baseline, use this skill to run TabPFN v2.5 on a fixed data split with CV folds. It generates out-of-fold predictions and a ready-to-submit file to accelerate leaderboard timing.

Yes, you can ensemble TabPFN predictions with gradient-boosted trees. The skill supports gating TabPFN outputs into GBT ensembles, allowing rapid feature probing, 1st-level calibration, and ensemble experiments for robust decision making.

Generate out-of-fold predictions by running TabPFN classification across predefined CV folds. The skill tracks CV vs LB relationships and saves artifacts, ensuring reproducible baseline metrics and valid submission formats for leaderboard submissions.

Before running the TabPFN classification baseline, you need prepared data and CV folds. The skill conforms to project conventions in tabpfn-core and references api-setup and data-requirements, meaning data exploration should be completed first.

Yes, the classification baseline supports rapid feature probing and threshold tuning. It applies TabPFN v2.5 to fixed data splits, enabling fast iteration on feature engineering and threshold adjustments before generating final submissions.

Use TabPFN to establish a fast, competitive baseline before investing in heavy gradient-boosted tree tuning. It accelerates leaderboard timing by producing immediate OOF predictions and submissions, which can then be gated into GBT ensembles for further gains.