TRivia-3B: Demo

Self-supervised Fine-tuning of Vision-Language Models for Table Recognition.