Even (very) noisy LLM evaluators are useful for improving AI agents(tensorzero.com)14 points by GabrielBianconi 2 days ago | 1 comment