Glossary · Term

multi-hop

← all terms

Definition

Plain language

A question you can only answer by chaining together several facts looked up one after another.

As stated in the literature

A question-answering regime requiring sequential retrieval where later queries depend on earlier answers; bridge questions in HotPotQA are the canonical example and resist parallel re-sampling.

Also called: multi-hop reasoning

Why it matters: Because each step depends on the previous answer, these questions can't be shortcut and test a system's ability to chain reasoning.

For example, answering 'what's the capital of the country where this author was born?' requires first finding the country, then looking up its capital.

Heard on the show

“What dies is the deliberate stuff — multi-hop reasoning, translation, and building abstract summaries of a situation.”
Episode 203 — The Thought a Model Doesn't Say — and the Lens That Reads It

Mentioned in 10 episodes

  1. 203
    The Thought a Model Doesn't Say — and the Lens That Reads It
  2. 200
    The One Mechanism That Turns Twenty AI Clones Into an Actual Team
  3. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
  4. 104
    How Making a Research Agent Smarter Quietly Makes It Leak Your Secrets
  5. 085
    Why Long-Context Models Might Need Compute, Not Capacity, Before Eviction
  6. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
  7. 030
    Why Your AI Agent Won't Stop Working — and Each Model Falls for a Different Trap
  8. 028
    Teaching a Model to Hire Copies of Itself: Recursive Agent Optimization
  9. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
  10. 011
    When RL Actually Teaches Agents Something New, And When It Doesn't

Related terms