RetGen: A Joint Framework for Retrieval and Grounded Text Generation Modeling


  • Yizhe Zhang Microsoft
  • Siqi Sun Microsoft
  • Xiang Gao Microsoft
  • Yuwei Fang Microsoft
  • Chris Brockett Microsoft
  • Michel Galley Microsoft
  • Jianfeng Gao Microsoft
  • Bill Dolan Microsoft



Speech & Natural Language Processing (SNLP), Machine Learning (ML)


Recent advances in large-scale pre-training such as GPT-3 allow seemingly high quality text to be generated from a given prompt. However, such generation systems often suffer from problems of hallucinated facts, and are not inherently designed to incorporate useful external information. Grounded generation models appear to offer remedies, but their training typically relies on rarely-available parallel data where information-relevant documents are provided for context. We propose a framework that alleviates this data constraint by jointly training a grounded generator and document retriever on the language model signal. The model learns to reward retrieval of the documents with the highest utility in generation, and attentively combines them using a Mixture-of-Experts (MoE) ensemble to generate follow-on text. We demonstrate that both generator and retriever can take advantage of this joint training and work synergistically to produce more informative and relevant text in both prose and dialogue generation.




How to Cite

Zhang, Y., Sun, S., Gao, X., Fang, Y., Brockett, C., Galley, M., Gao, J., & Dolan, B. (2022). RetGen: A Joint Framework for Retrieval and Grounded Text Generation Modeling. Proceedings of the AAAI Conference on Artificial Intelligence, 36(10), 11739-11747.



AAAI Technical Track on Speech and Natural Language Processing