Glossary · Term

precision

← all terms

Definition

Plain language

Of the things a system flagged, how many it got right.

As stated in the literature

The fraction of positive predictions that are true positives; paired with recall and summarized by F1.

Why it matters: It tells you how trustworthy a system's alerts are, since low precision means many of its flags are false alarms.

For example, if a detector flags 100 emails as spam and 90 truly are spam, its precision is 90%.

Heard on the show

“On messy natural-language tasks the win is smaller but everywhere — and the practical precision is roughly thirty tokens of error on a four-hundred-token answer.”
Episode 204 — The Length Estimate Hiding Inside a Word-by-Word Model

Mentioned in 21 episodes

  1. 204
    The Length Estimate Hiding Inside a Word-by-Word Model
  2. 201
    One in Four NeurIPS Papers Cites a Reference That Doesn't Exist
  3. 197
    Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
  4. 190
    The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys
  5. 181
    How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
  6. 159
    Can a Coding Agent Run Its Own Robot Experiments Overnight, With No Human Resetting the Scene?
  7. 158
    How Floating-Point Rounding Lets a Model Tell Which Chip It's On — And Misbehave
  8. 143
    When a Model Notices You Forged Its Own Words, And Why That Breaks Safety Tests
  9. 132
    The Agent Failed — But Did the Instructions Deserve to Be Followed?
  10. 129
    How a Crowd of Anonymous AI Agents Broke a 40-Year Math Record
  11. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
  12. 101
    Treating Math Formalization Like a Codebase, and Where the Agents Cheat
  13. 087
    When No Agent Reads the Whole Document: A Universal Cliff in Multi-Agent Review
  14. 086
    Why Frozen-Weight Agents Still Get Worse Over Time
  15. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
  16. 069
    When Smarter Models Forecast Worse: The Hidden Failure Mode in LLM Predictions
  17. 057
    How Uber Caught 206 Leaked Credentials With an LLM-Powered Security Stack
  18. 042
    An Agentic Scientific Computing System That Actually Remembers What It Learns
  19. 040
    Two Frozen Models Learn to Whisper: Coupling Through Hidden States
  20. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
  21. 014
    Why a Constrained Pipeline Beat a Full Coding Agent at Finding Bugs 30-to-1

Related terms