Скачать или смотреть Retrieval Augmented Generation (RAG)

The paper introduces Retrieval-Augmented Generation (RAG) models, which combine a pre-trained parametric sequence-to-sequence model (like BART) with an external non-parametric memory (a dense vector index of Wikipedia passages) accessed via a neural retriever.
These models are trained end-to-end and set the state of the art on multiple open-domain Question Answering tasks while generating responses that are more factual, specific, and diverse than purely parametric models.

Citation:
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W-T., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. (arXiv:2005.11401).