trends.barnas.net

benchmark · first seen 5 Nov 2026

WMDP benchmark

Also: Weapons of Mass Destruction Proxy

Multiple-choice benchmark of proxy questions on hazardous knowledge used for safety evaluation.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

safety refusal in language modelslogit lens