LESSWRONG
Wikitags
LW

Sandbagging (AI)

Written by Raemon last updated 27th Mar 2025

Sandbagging is when an AI system pretends to be less capable during training/evaluation.

Posts tagged Sandbagging (AI)

3

43Notes on countermeasures for exploration hacking (aka sandbagging)

ryan_greenblatt

19d

6

2

56The “no sandbagging on checkable tasks” hypothesis

2y

14

2

41Automated Researchers Can Subtly Sandbag

gasteigerjo, Akbir Khan, Sam Bowman, Vlad Mikulik, Ethan Perez, Fabien Roger

17d

0

1

84[Paper] AI Sandbagging: Language Models can Strategically Underperform on Evaluations

Teun van der Weij, Felix Hofstätter, Ollie J, Sam F. Brown, Francis Rhys Ward

10mo

10

1

45An Introduction to AI Sandbagging

Teun van der Weij, Felix Hofstätter, Francis Rhys Ward

1y

13

1

23How to mitigate sandbagging

Teun van der Weij

20d

0

1

15Won't vs. Can't: Sandbagging-like Behavior from Claude Models

Joe Benton, Zachary Witten

2mo

1