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

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

10 points by GabrielBianconi 2 days ago | discuss