1 paper · 1 filter
Joy Mahapatra, Utpal Garain
Large Language Models (LLMs) have shown exceptional performance across various Data-to-Text Generation (DTG) tasks. However, generating factually consistent text in DTG remains cha…