This post is a brief summary about the paper that I read for my study and curiosity, so I shortly arrange the content of the paper, titled Self-RAG - Learning to Retrieve, Generate, and Critique through Self-Reflection (Asai et al., arXiv 2023), that I read and studied.

They was saying that RAG (Retrieval-Augmented Generation) is ad-hoc approach to augments LM with retrieval of relevant knowledge to decreases producing factual inaccuracies.

But, it doesn’t take into account whether retireval is necessary or passage are relevant and then it diminishes LM versaltility or can lead to unhelpful response generation.

So, they propose a new framework called Self-Relective Retrieval-Augmented Generation (SELF-RAG).

Asai et al. arXiv 2023

For detailed experiment and explanation, refer to the paper, titled Self-RAG - Learning to Retrieve, Generate, and Critique through Self-Reflection (Asai et al., arXiv 2023)

Reference