Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku
Anthropic
partial detail 19 method fields published
Evaluation
Refusal evaluations (Wildchat and XSTest)
refusal rates on toxic prompts (should refuse) and incorrect refusal rates on non-toxic prompts (should not refuse), using Wildchat and XSTest datasets
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 3.5 Sonnet
Anthropic
partial detail 15 method fields published
See how this test was run
Claude 3.5 Haiku, Claude 3.5 Sonnet (New)
Anthropic
partial detail 15 method fields published