learn.ifs.name

Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection

Asai et al., 2024, Self-RAG (The Twelfth International Conference on Learning Representations (ICLR 2024))

Subscriber deck — free during early access

1 cards in “4.3 Experimental Settings” · whole deck

4.3 Experimental Settings

  1. On what data and models was Self-RAG trained?
    150K instruction–output pairs (from Open-Instruct and knowledge-intensive datasets), with Llama 2 7B and 13B as generators, Llama 2 7B as critic, and Contriever-MS MARCO retrieving up to ten documents.
    • Asai et al., 2024, Self-RAG (The Twelfth International Conference on Learning Representations (ICLR 2024)), §4.3

Source

  • Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil and Hannaneh Hajishirzi. 2024. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. In The Twelfth International Conference on Learning Representations (ICLR 2024). https://arxiv.org/abs/2310.11511