liveBook · Manning

Chain-of-thought

As explained in Retrieval Augmented Generation, The Foundational Ideas

Multi-step reasoning coordination is the organization of subproblem solving, information gathering, and result synthesis across iterative reasoning cycles.

Multi-step reasoning coordination organizes complex work into subproblems, directs repeated information gathering, and synthesizes results across iterative retrieval and generation cycles. It supports workflows where information needs evolve during processing rather than remaining fixed at query time. section 1.5.4

A coordinated system can:

  • Generate searches from predicted information needs or detected knowledge gaps. section 1.5.4
  • Choose among a vector index, SQL database, and web search for different subquestions. section 1.5.4
  • Assess information quality and trigger correction through confidence scoring, factual verification, or gap detection. section 1.5.4

Research agent

A research agent plans a sequence of searches, runs them, evaluates the results, and decides whether another search is needed before synthesizing an answer. section 1.5.4

How Retrieval Augmented Generation, The Foundational Ideas covers it

How deep it goes
Defines iterative decomposition, retrieval, validation, and synthesis, but does not specify a complete implementation.
Examples
Uses research agents, tool-selecting systems, and ScholarCopilot to illustrate iterative retrieval and specialized information gathering.
What you can do after
Evaluate multi-step processes with monitoring, confidence scoring, factual verification, and gap detection.
How it connects
Connects query formulation, anticipatory retrieval, reactive retrieval, self-monitoring, factual verification, and multi-hop reasoning.

Also explained in

Compare how 21 books explain chain-of-thought →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Retrieval Augmented Generation, The Foundational Ideas. The section links point to the parts of the book it draws on.