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

by GabrielBianconi | View on Hacker News