recall

Definition

Plain language

Of the things a system should have caught, how many it actually caught.

Why it matters: It measures how much a system misses, which is critical for safety monitors where letting a real attack slip through is costly.

For example, if there were 100 spam emails and the filter caught 80 of them, its recall is 80%.

Heard on the show

“The correct answer to those questions saturates the training data, so a direct question tests recall, not judgment.”

Mentioned in 33 episodes

1. 240
   Frontier Models Designed Follow-Ups To Fraudulent Papers 93% Of The Time
2. 238
   How a Cheap Model Reads the Flagship's Secret Reasoning Aloud
3. 233
   Why a Model Can Grade an Answer But Not Write the Answer Key
4. 212
   The Fact Was in the Wrong Drawer: Why Fine-Tuned Models Can't Reason With What They Know
5. 197
   Twin Problems Suggest AI Reasoning Gains Are Mostly Better Fact Recall
6. 192
   A 32B Open Model Matched Frontier Systems By Learning to Take Notes
7. 191
   How One Researcher Beat GPT-5.2 and Gemini 3 by Judging Their Answers, Not Improving Them
8. 187
   An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
9. 186
   How a Frozen Model Went From 2% to 77% on Physics Puzzles — Without Retraining
10. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
11. 180
    The Bug Where Smart Assistants Read a Fact and Still Forget It
12. 169
    Why Better Bug Reports Can Make AI Coding Agents Worse
13. 145
    Building Forgetting Into a Language Model With One Extra Line of Code
14. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
15. 114
    Agents That Rewrite Their Own Weights Instead of Just Taking Notes
16. 108
    The Reasoning Cliff: Why Thinking Longer Makes Models Worse at Exact Step-by-Step Tasks
17. 102
    How to Catch an AI Attack That No Single Conversation Reveals
18. 100
    How a Prompt Wrapper Lets a Frontier Model Play Poker Like an Expert
19. 089
    When AI-Written Papers Read Well But the Evidence Underneath Is Broken
20. 086
    Why Frozen-Weight Agents Still Get Worse Over Time
21. 085
    Why Long-Context Models Might Need Compute, Not Capacity, Before Eviction
22. 077
    Reading a Model's Confidence Curve to Decide When Chain-of-Thought Is Worth It
23. 074
    How a Fifteen-Hundred-Dollar Training Run Matched Llama and Gemma on Reasoning
24. 062
    Treating Hallucinations as Exploits: A Gate-Based Architecture for Agent Safety
25. 057
    How Uber Caught 206 Leaked Credentials With an LLM-Powered Security Stack
26. 040
    Two Frozen Models Learn to Whisper: Coupling Through Hidden States
27. 038
    How LLMs Get Persuaded: One Attention Head, A Tetrahedron, And A Single Dial
28. 037
    Why Hallucination Detectors Miss Stale Facts: A Geometric Story About What Models Know But Don't Say
29. 036
    Sparse Attention Was the Wrong Frame. Treat It as Geometry Instead.
30. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval
31. 032
    A Sticky-Note for Every Layer: Letting Transformers Remember What They Were Just Thinking
32. 031
    When Your AI Assistant Won't Let Go of Old Facts About You
33. 023
    Why a Small Agent Confidently Overwrites Memories It Doesn't Understand

Related concepts

Related terms