YFarmX

Model evaluation for dangerous capabilities

AI

Model evaluation for dangerous capabilities: Structured testing of whether a model meaningfully helps with cyber-attacks, biological weapons or autonomous replication, before release.

Related terms

Browse the full glossary →