Improving Factuality in LLMs via Inference-Time Knowledge Graph Construction
arXiv:2509.03540v3 Announce Type: replace-cross
Abstract: Large Language Models (LLMs) often struggle with producing factually consistent answers due to limitations in their parametric memory. Retrieval-Augmented Generation (RAG) paradigms mitigate th…