Research question
How should defences against adversarial examples be evaluated?
Key findings
- A checklist of common evaluation pitfalls and required adaptive attacks - Recommends threat-model clarity, sanity checks and releasing code
On Evaluating Adversarial Robustness
The field's consensus checklist for evaluating adversarial defences, written by leading researchers after years of broken defences.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.