HN.zip

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

11 points by GabrielBianconi - comments