Dr. Aleksandar Tošić and Domen Vake, InnoRenew CoE researchers, together with Jernej Vičič from the University of Primorska’s Faculty of Mathematics, Natural Sciences and Information Technologies, recently published an article »Bridging the Question–Answer Gap in Retrieval-Augmented Generation: Hypothetical Prompt Embeddings,« in IEEE Access.

Retrieval-Augmented Generation (RAG) systems synergize retrieval mechanisms with generative language models to enhance the accuracy and relevance of responses. However, bridging the style gap between user queries and relevant information in document text remains a persistent challenge often addressed by runtime solutions such as Hypothetical Document Embeddings (HyDE). To address these challenges, authors propose Hypothetical Prompt Embeddings (HyPE), a framework that shifts the generation of hypothetical content from query time to the indexing phase. Results showed that HyPE can improve retrieval context precision.