Primer
  • Factored Cognition Primer
  • Intro
    • Factored Cognition
    • Before We Start
  • Chapters
    • Hello World
    • Question Answering
      • Q&A without context
      • Q&A about short texts
    • Debate
      • Representing debates
      • From debates to prompts
      • The debate recipe
    • Long Texts
      • Loading paper text
      • Finding relevant paragraphs
      • Answering given paragraphs
    • Amplification
      • Asking subquestions
      • Answering subquestions
      • One-step amplification
      • Recursive amplification
    • Verifiers
      • Checking answers
      • Checking reasoning steps
    • Tool Use
      • Web search
      • Interpreters
    • Deduction
      • Chain of Thought
    • Action Selection
      • One-shot action selection
      • Iterative action selection
    • Amplification Revisited
  • Appendix
    • What’s next?
  • Links
    • We’re Hiring
    • Our Slack Community
    • ICE on Github
Powered by GitBook
On this page
Edit on GitHub
  1. Chapters

Action Selection

So many things a model could do

PreviousChain of ThoughtNextOne-shot action selection

Last updated 2 years ago

We’ve seen different cognitive actions that a model can take to answer a question, including:

  1. Just answer the question directly.

  2. Run a debate if it’s a pro/con question.

  3. Search a long text for relevant information.

  4. Decompose the question into subquestions.

  5. Run a web search using Google.

  6. Run a computation in Python.

  7. Write out reasoning steps.

In this chapter, we’ll use a model to choose which of these to run. We’ll look at two cases:

  1. : Just choose a single action.

  2. : Given the results of actions so far, choose the next action.

One-shot action selection
Iterative action selection