Building Effective AI Agents
… This tends to perform better than having the same LLM call handle both guardrails and the core response. Automating evals for evaluating LLM performance, where each LLM call evaluates a different aspect of the model’s performance on a given prompt. …