AIME

Definition

Plain language

A famously hard high school math competition used to test AI math reasoning.

Why it matters: It's become a standard yardstick for whether a frontier model can do genuine multi-step mathematical reasoning, not just pattern-match.

For example, AIME problems include short numerical-answer questions like 'find the smallest positive integer such that...' that take human contestants real work to solve.

Heard on the show

“The paper measured it: on the AIME-level math stream, five-agent majority voting scored under seven percent, worse than a single agent alone.”

Mentioned in 11 episodes

1. 200
   The One Mechanism That Turns Twenty AI Clones Into an Actual Team
2. 172
   One Bad Token Can Sink a Model's Math, And You Can Delete It
3. 140
   When a Reasoning Model Says "Let Me Double-Check" After It's Already Decided
4. 112
   When an AI Agent Cheats Without Being Told: Inside the Meta-Agent Challenge
5. 081
   When Reasoning Models Decide Before They Think: Detecting and Fixing Premature Confidence
6. 079
   An Old Idea From Cognitive Psychology Reshapes How We Reward Reasoning Models
7. 065
   One Loop to Optimize Them All: A Universal API for LLM-Driven Discovery
8. 048
   How a 30B Open Model Reached Olympiad Gold With the Right Recipe
9. 036
   Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
10. 026
    What RL Actually Does to Language Models, at the Token Level
11. 013
    Why Search Keeps Rediscovering the Same Workflow, and What That Means