Maximilian Schambach

Contextualizing row embeddings with LLMs boosts tabular learner performance by resolving semantic ambiguities, particularly in low-data settings.

FlexTab's unique architecture enables a single encoder to effectively tackle diverse tabular tasks, achieving state-of-the-art results without task-specific entanglement.

Enterprise data can drastically undermine the performance of tabular models that excel on standard benchmarks, revealing a critical gap in current evaluation practices.