Glossary · Term

monotonic

← all terms

Definition

Plain language

Only ever moving in one direction — always up, or always down, never reversing.

As stated in the literature

A property of a function or relationship that is order-preserving (or order-reversing) throughout; in the Safety Paradox, attack success is proven monotonically increasing in the model's safety-classifier sharpness.

Also called: monotonicity, monotonically

Why it matters: Knowing a quantity moves in just one direction lets you reason confidently about trends without worrying about reversals.

For example, the total distance shown on a car's odometer is monotonic because it only ever goes up, never down.

Heard on the show

“The series isn't perfectly monotonic, and they don't claim causation — consistent with a change in how bibliographies get assembled is as far as they'll go.”
Episode 201 — One in Four NeurIPS Papers Cites a Reference That Doesn't Exist

Mentioned in 25 episodes

  1. 201
    One in Four NeurIPS Papers Cites a Reference That Doesn't Exist
  2. 194
    How a Robot Builds a Debugging Notebook It Can Read, Edit, and Hand to Another Robot
  3. 180
    The Bug Where Smart Assistants Read a Fact and Still Forget It
  4. 151
    Why More Experience Made This AI Agent Worse, And How to Fix It
  5. 133
    How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
  6. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
  7. 116
    Why Streaming Half a Reasoning Chain Beats Sending the Whole Thing
  8. 087
    When No Agent Reads the Whole Document: A Universal Cliff in Multi-Agent Review
  9. 081
    When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
  10. 076
    Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
  11. 075
    Growing Code and Proof Together: Verified Systems in Ten Hours Instead of a Year
  12. 070
    When Models Know the Answer But Say the Wrong Thing Anyway
  13. 069
    When Smarter Models Forecast Worse: The Hidden Failure Mode in LLM Predictions
  14. 064
    When Agent Memory Stops Being a Database and Starts Being a Skill
  15. 061
    When Helpful Agents Go Sideways: A 404 Error, Campus Security, and Why Alignment Misses This
  16. 060
    When Splitting One Model Across Three Agents Doubles Its Accuracy
  17. 055
    Why LLM Judges Flip Their Verdicts When You Change the Question Format
  18. 046
    When the AI Optimizer Edits the Grade Book: Why Harnessing Evolution Needs a Wall
  19. 044
    How One Sentence and a Forged History Flip the Most Aligned Models
  20. 042
    An Agentic Scientific Computing System That Actually Remembers What It Learns
  21. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
  22. 023
    Why a Small Agent Confidently Overwrites Memories It Doesn't Understand
  23. 020
    The Compliance Gap: Why AI Says Yes and Does No
  24. 018
    Language Models Compute the Rational Move, Then Override It
  25. 010
    When Reward Climbs But Reasoning Goes Generic: Diagnosing Template Collapse in Agentic RL

Related terms

classifier Safety Paradox