TabFM-Auto: Self-Evolving Pipelines for Tabular Foundation Models

The paper introduces TabFM-Auto, a method that pairs the TabFM tabular foundation model with an LLM agent to iteratively evolve surrounding data pipelines. Guided by dataset metadata and validation feedback, the agent refines cleaning, feature engineering, and post-processing to reduce model error. Across all 51 datasets of the TabArena benchmark, five TabFM-Auto configurations claim the top five overall positions, with the best configuration increasing TabFM's Elo from 1785 to 2013. Furthermore, the discovered pipelines transfer directly to other frozen tabular foundation models, boosting their Elo by +69 to +143 without additional search.

Five TabFM-Auto configurations take the top five overall positions across all 51 datasets of the TabArena benchmark.

The best configuration raises TabFM from 1785 to 2013 Elo, and discovered pipelines transfer to other frozen tabular foundation models with no further search (+69 to +143 Elo).

Five TabFM-Auto configurations take the top five overall positions across all 51 datasets of the TabArena benchmark.