AI RESEARCH
The Illusion of Generalization in Tabular Language Models
arXiv CS.LG
•
ArXi:2602.04031v2 Announce Type: replace Tabular Language Models (TLMs) have been claimed to achieve strong generalization for tabular prediction. We conduct a systematic re-evaluation of Tabula-8B as a representative TLM, utilizing 165 datasets from the UniPredict benchmark. Our investigation reveals three findings. First, binary and categorical classification achieve near-zero median lift over majority-class baselines and strong aggregate performance is driven entirely by quartile classification tasks.