RAG (Retrieval-Augmented Generation) combines an external retriever and a generative model to ground LLM outputs by marginalizing over retrieved documents, P(y|x)=∑_{i=1}^{K} P_ret(z_i|x) P_gen(y|x,z_…
Tag: RAG
RAG (Retrieval-Augmented Generation) combines an external retriever and a generative model to ground LLM outputs by marginalizing over retrieved documents, P(y|x)=∑_{i=1}^{K} P_ret(z_i|x) P_gen(y|x,z_…