Public demo

Retrieval-Augmented Generation

A source-grounded theme page on retrieval-augmented generation: original RAG, Self-RAG, retrieval control, and evidence support.

Theme overview

This theme follows retrieval-augmented generation from the original retrieval-plus-generator architecture to models that decide when and how to retrieve. It is the closest demo theme to SAV's core product idea: writing with source support from a corpus.

Takeaways

• RAG is not just a model pattern; it is a trust workflow problem.
• Retrieval helps only when the user can inspect the source and decide whether it supports the claim.
• SAV should present itself as evidence-supported research writing, not as a generic RAG chatbot.

Included papers

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection

Evidence examples

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks: Integrating hybrid parametric/non-parametric memory and retrieval modules enhances pre-trained generative language models for broader NLP task capabilities.

  Here, we bring hybrid parametric and non-parametric memory to the "workhorse of NLP," i.e. sequence-to-sequence (seq2seq) models.

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks: The paper proposes generating Jeopardy questions to evaluate RAG's generation abilities in a non-QA setting, as it is a more demanding task than standard open-domain QA.

  To evaluate RAG's generation abilities in a non-QA setting, we study open-domain question generation.

• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection: This work introduces Self-Reflective Retrieval-augmented Generation (SELF-RAG) to improve an LLM's generation quality, including its factual accuracy without hurting its versatility, via on-demand retrieval and self-reflection.

  This work introduces Self-Reflective Retrieval-augmented Generation (SELF-RAG) to improve an LLM's generation quality, including its factual accuracy without hurting its versatility, via on-demand retrieval and self-reflection.

• Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection: Approaches are described to enforce control during the inference process to meet distinct task objectives.

  In this section, we describe approaches to enforce control to meet these distinct objectives during the inference process.