Even (very) noisy LLM evaluators are useful for improving AI agents

tensorzero.com

8 points by GabrielBianconi 2 days ago