AI
TabFM-Auto: Self-Evolving Pipelines for Tabular Foundation Models
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 28/09/2026, 20:00:00 UTC
Briefing created: 30/09/2026, 04:34:33 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 30/09/2026, 04:18:24 UTC.
Briefings
TabFM-Auto combines a tabular foundation model with an evolving machine learning agent to automatically refine data pipelines. Across 51 TabArena datasets, five configurations achieved top five rankings, boosting TabFM's Elo rating from 1785 to 2013. On MLE-Bench, the system ranks first among all MLE agents on 8 competitions. The approach transfers to other frozen models, improving their performance by 69 to 143 Elo without further search. Uncertainty remains regarding the specific agent configurations and the exact transferability to new models.
Key points
- TabFM-Auto pairs a tabular foundation model with a language model agent to iteratively refine data pipelines guided by metadata and validation feedback.
- Across 51 TabArena datasets, five configurations achieved top five rankings, raising TabFM's Elo rating from 1785 to 2013.
Context and limitations
Coverage is limited to 51 TabArena datasets and 8 MLE-Bench competitions; results may not generalize to other frozen tabular foundation models or new datasets.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 30/09/2026, 04:34:33 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeidnxa3jl3bgnwx4e5ker6p6rqel5ppl7uivad6wbek6mzyrylniou