Why Large Language Models Fail at Tabular Prediction

Marta Garnelo, Wojciech M. Czarnecki

Large language models (LLMs) have become the default tool for a remarkable range of tasks, yet they have had conspicuously little success at one of the most common machine learning…