tabpfn-regress

What problem does it solve?

TabPFN-based regression benchmarks traditionally require manual, multi-step pipelines to establish baselines and generate submissions for Kaggle-like competitions. This Skill provides an end-to-end regression baseline, including OOF generation, first submission, and subsequent optimization with ensemble methods and post-processing rules.

Core Features & Use Cases

• Regression baseline: a TabPFN v2.5 baseline with cross-validated out-of-fold predictions.
• Ensembling & post-processing: gradient-boosting trees ensembles, clipping, target transforms, rank blending, and a robust submission workflow.
• Use Case: After tabpfn-explore has prepared data and folds, produce a reproducible submission and CV log for leaderboard comparison.

Quick Start

Run the TabPFN regression baseline to generate OOF predictions and a first submission after data preparation and fold generation.