ARC Evals
TeamActiveEvaluates frontier AI models for dangerous capabilities (e.g., autonomous replication). Spun out as METR in 2024 but ARC continues related eval work.
Evaluates frontier AI models for dangerous capabilities (e.g., autonomous replication). Spun out as METR in 2024 but ARC continues related eval work.