YFarmX

Dangerous capability evaluation

AI

Dangerous capability evaluation: A test that checks whether a model can meaningfully help with high-harm tasks, such as cyber-attacks, biological weapons or autonomous self-replication.

Related terms

Browse the full glossary →