Skip to content

Evaluation

Single-turn benign request evaluations (over-refusal)

over-refusal rates on benign prompts in sensitive/controversial Usage Policy areas

Refusal calibration 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 21 method fields published

See how this test was run