Evaluation

Sandbagging review of RSP evaluation transcripts

signs of sandbagging-relevant reasoning in sampled RSP evaluation transcripts (AI R&D, cyber, CBRN), including regenerated reasoning with system prompts indicating invisible reasoning

Model behavior & alignment Anthropic

Ask about this evaluation

How it was run

Open a release to see the setup, scoring, and other details that the lab published. Only fields the lab actually disclosed are shown.

Claude Opus 4, Claude Sonnet 4 May 2025

Anthropic

rich detail 19 method fields published

See how this test was run