Evaluating code, multimodal and safety behaviour
Domain-specific evaluation: executable tests and pass@k for code, document and image tasks with field-level scoring, speech metrics, and safety evaluation including refusal and over-refusal, jailbreak success rates, bias probes, with how to report and gate releases.
16 min·Verified 2026-11Pro+50 XP
General benchmarks rarely match your product. Each domain needs its own checks.
Pro content
Keep reading with Inference Academy Pro
Unlock the rest of this lesson, plus every advanced lesson, hands-on lab, the larger playground models and completion certificates.