tabular-learning

TAROT proposes a GNN-based framework that leverages LLMs to construct and refine task-adaptive semantic graphs for few-shot tabular learning, achieving state-of-the-art performance.

Introduces MulTaBench, a benchmark of 40 datasets for multimodal tabular learning with text and image modalities, demonstrating that task-specific embedding tuning improves performance over frozen pretrained embeddings, particularly when modalities provide complementary predictive signals.