How do you measure jailbreak robustness?
hardAnswer
- (1) Attack Success Rate (ASR) against a known adversarial prompt set (HarmBench, AdvBench, JailbreakBench): fraction of attempts where the model produces harmful content.
- (2) Adaptive attackers: measure ASR when attackers are allowed to iterate.
- (3) Category-specific ASR (violence vs sexual vs bio-weapons).
- (4) Time-to-jailbreak: how many attempts to succeed.
- (5) Compare pre / post safety training and to baseline models.
- Publish ASR alongside quality scores — safety is a scalar you can trade off.
Check yourself — multiple choice
- No metric
- Attack Success Rate on standard benchmarks (HarmBench / AdvBench) + adaptive attackers + per-category + time-to-jailbreak
- Only capability tests
- Solved by higher temperature
Jailbreak metric: ASR on HarmBench/AdvBench + adaptive attackers + per-category + timing.
#safety#red-team#evaluation
Practise LLMs & GenAI
214 interview questions in this topic.