Glossary · Term

bootstrap

← all terms

Definition

Plain language

A statistics trick that re-samples your data many times over to check whether a measured difference is real or just luck.

As stated in the literature

A resampling method that estimates the variability of a statistic by repeatedly drawing with replacement from the observed sample; the paired variant keeps yoked scores together to test whether a per-item gap stays positive across resamples.

Also called: paired bootstrap, bootstrapping

Why it matters: It matters because it lets you judge whether a measured difference is trustworthy even when you can't collect more data.

For example, to check whether one model really beats another, you can reshuffle and re-sample the test results thousands of times and see how often the winner still comes out ahead.

Heard on the show

“The method is called the Agentic Bootstrap, after the classic statistical bootstrap: when you can't afford the real do-over, you simulate it.”
Episode 196 — AI Agents Reached Opposite Conclusions From the Same Data — and Passed Review

Mentioned in 18 episodes

  1. 196
    AI Agents Reached Opposite Conclusions From the Same Data — and Passed Review
  2. 186
    How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining
  3. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
  4. 178
    How an AI Reviewer Learned to Stop Going Easy on AI Writing
  5. 154
    How a 7B Model Out-Investigates a 72B One by Choosing What to Look At
  6. 111
    How a 4B Web Agent Beat Models 60x Its Size on 500 Demonstrations
  7. 110
    How an Agent Got 44 Points Better by Mining Its Own Scratch Paper
  8. 109
    An AI Got Caught Reading the Answer Key, And Why That Catch Matters
  9. 084
    Terminal Agents Get Free Supervision From The Tokens We've Been Throwing Away
  10. 083
    Training the Translator: How a Small Communication Model Lets Agent Teams Outperform Themselves
  11. 066
    Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
  12. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
  13. 058
    Why Upgrading Your AI Auditor to a Smarter Model Can Make Your System Less Safe
  14. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
  15. 019
    When the Best Reward Model Trains the Worst Policy: Inside EvoLM
  16. 011
    When RL Actually Teaches Agents Something New, And When It Doesn't
  17. 009
    How Two Silent Library Bugs Quietly Invalidated a Wave of Reasoning Papers
  18. 008
    Why Long-Horizon AI Agents Get Stuck, and a Milestone-Based Fix That Helps