Self-RAG is a new framework to train an arbitrary LM to learn to retrieve, generate, and critique to enhance the factuality and quality of generations, without hurting the versatility of LLMs.
Primary LanguageJupyter Notebook
No issues in this repository yet.