hallucination

Definition

When an AI confidently states something that isn't true.

A failure mode in which a language model generates content that is fluent and confident but factually incorrect or unsupported.

Also called: hallucinations, hallucinate, hallucinated, hallucinating

Mentioned in 16 episodes

1. 076
   Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
2. 072
   A Robot Made Graphene Without Help, And Caught Itself Hallucinating
3. 070
   When Models Know the Answer But Say the Wrong Thing Anyway
4. 067
   An AI Just Solved a 1996 Erdős Problem—and the Simplest Agent Won
5. 066
   Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
6. 062
   Treating Hallucinations as Exploits: A Gate-Based Architecture for Agent Safety
7. 059
   Firefly's Inversion: Building Verified Tool-Call Training Data by Working Backward
8. 055
   Why LLM Judges Flip Their Verdicts When You Change the Question Format
9. 052
   An Old Reinforcement Learning Tradeoff Sneaks Back Into LLM Agents
10. 041
    When the Iteration Teaches the Model to Skip the Iteration
11. 039
    When Smarter Agents Get Fooled by Three Extra Nodes in a Database
12. 037
    Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
13. 029
    Why Forty-Eight Percent on FrontierMath Isn't the Real Story in DeepMind's New Math Paper
14. 025
    The Missing Gradient Term That Predicts Sycophancy in RLHF
15. 017
    When the Agent Grades Its Own Homework: A Brutal New Benchmark for AI Workers
16. 014
    Why a Constrained Pipeline Beat a Full Coding Agent at Finding Bugs 30-to-1