Encyclopedia Evalica / Datasets / Adversarial examples
Adversarial examples
/a.dver'seh.ree.uhl ih'gza.mpuhlz/Test cases designed to break the system (prompt injection, tricky edge cases, ambiguous inputs) to reveal failure modes. They often complement "happy path" test sets. (noun)
“We added adversarial examples to make sure the agent resists prompt injection.”
Related Datasets terms
From the docs
Get started with Evals
Braintrust is the observability platform for agents in production. By actively applying intelligence to agent traces and automatically surfacing critical patterns, Braintrust helps teams at Notion, Stripe, Box, OpenAI, and Cloudflare ship quality agents at scale.
Start building