Adding knowledge to open-source LLMs

While Large Language Models are remarkably powerful, their knowledge is typically limited to the data available during training. This makes their applications challenging when business contexts, or cultural nuances change or are not well represented in their original training data. In many cases, in-context learning techniques such as retrieval-augmented generation (RAG) can help to bridge this gap by providing models with relevant information at runtime. However, when a model lacks foundational understanding of domain-specific knowledge, use case, or local cultural context, even advanced retrieval methods may fail. In this session, NVIDIA experts will explain how to enrich language models with new knowledge, expanding their capabilities in specialized business, engineering, or scientific domains, and adjusting adaptation to new languages, cultures, and values.