Demystifying evals for AI agents \ Anthropic
Demystifying evals for AI agents