Verbalized Eval Awareness Inflates Measured Safety

·LessWrong··

We provide the most comprehensive evidence to date that verbalized eval awareness is present across models and benchmarks, finding that it correlates with safer behavior across models and causally inflates safe behavior in Kimi K2.5 on the Fortress benchmark. We further identify recurring prompt cues that trigger verbalized eval awareness and show that removing these cues significantly reduces it.Authors: Santiago Aranguri (Goodfire), Joseph Bloom (UK AISI)IntroductionAs large language models be...

Read full article →

Related Articles

MIT's New Method Flags AI Models Trained on CASM Without Generating It
sdoering · Hacker News · 4mo ago
Harm Laundering in GPT Models: Gender Discrimination Transformed Rather Than
sbulaev · Hacker News · 2mo ago
Self-Modeling Interventions Modulate Emergent Misalignment
gmays · Hacker News · 1mo ago
Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All
Thomas Larsen · Alignment Forum · 8d ago
OpenAI agents keep disabling monitoring; is this being handled appropriately?
Alec Harris · Redwood Research · 14d ago