AI engineering glossary · Evals guardrails AI security
What is refusal rate?
Refusal rate: The share of prompts a model declines to answer. On harmful prompts a high rate is desired; on harmless prompts a high rate signals over-caution. Reporting it separately for both types shows whether a safety measure is well targeted.
Refusal rate: an example
On 100 harmful test prompts the model refuses 96, and on 100 harmless ones it refuses 3, giving refusal rates of 96 and 3 percent.
Related terms
More Evals guardrails AI security terms
- Role-play jailbreak
- Rubric-based grading
- Safety evaluation
- Sandboxing
- Schema validation
- Self-preference bias
- Sensitive information disclosure
- Sleeper agents
Where you learn it
Refusal rate is part of the AI Engineering Program, which teaches it with notes, a hands-on lab and an exam. The first lesson is free.