The Research Library connects handbook observations to public evidence. Each record separates what the source reports from why it matters and how the handbook interprets it.
Canonical evidence objects
information retrieval
Public sources only. The records are intentionally concise so future chapters, topic search, and Knowledge CI/CD integrations can reuse the same model.
Meta AI2020
Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
Lewis and colleagues describe a model that combines retrieved material with generation for knowledge-intensive language tasks.
What the source reports
The paper presents retrieval-augmented generation for knowledge-intensive NLP tasks.
It makes retrieval a distinct contribution to a language-model workflow.
Why it matters here
Chapter 4 explains what RAG contributes: bringing relevant retrieved material into a generative workflow.
Guided Intelligence interpretation
The source does not endorse Guided Intelligence. It helps distinguish retrieval and generation from the separate work of preserving governed reusable understanding.
Dense Passage Retrieval for Open-Domain Question Answering
Karpukhin and colleagues present dense passage retrieval as an approach to retrieving relevant passages for open-domain question answering.
What the source reports
The paper examines dense retrieval for question answering.
It separates retrieval of relevant passages from the later task of producing an answer.
Why it matters here
Chapter 4 distinguishes the valuable work of finding relevant information from deciding whether it is current, applicable, authorized, and reusable.
Guided Intelligence interpretation
The source does not endorse Guided Intelligence. It supports the chapter's careful distinction between retrieval capability and governed organizational understanding.
