Evaluation
Conversation termination with simulated users (Claude Opus 4)
Claude's preference to opt out of distressing conversations with abusive simulated users; percentage of conversations terminated
Model behavior & alignment
Anthropic
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
Anthropic
rich detail
19 method fields published
See how this test was run