https://blog.qife122.com/tags/%E9%80%92%E5%BD%92%E9%A1%BE%E9%97%AE/ Recent content in 递归顾问 on 办公AI智能小助手 Hugo zh-cn qife Tue, 25 Nov 2025 13:50:29 +0800 https://blog.qife122.com/p/%E4%BD%BF%E7%94%A8spring-ai%E6%9E%84%E5%BB%BAllm%E8%87%AA%E6%88%91%E8%AF%84%E4%BC%B0%E7%B3%BB%E7%BB%9F%E5%9F%BA%E4%BA%8E%E9%80%92%E5%BD%92%E9%A1%BE%E9%97%AE%E7%9A%84llm-as-a-judge%E5%AE%9E%E7%8E%B0/ Tue, 25 Nov 2025 13:50:29 +0800 https://blog.qife122.com/p/%E4%BD%BF%E7%94%A8spring-ai%E6%9E%84%E5%BB%BAllm%E8%87%AA%E6%88%91%E8%AF%84%E4%BC%B0%E7%B3%BB%E7%BB%9F%E5%9F%BA%E4%BA%8E%E9%80%92%E5%BD%92%E9%A1%BE%E9%97%AE%E7%9A%84llm-as-a-judge%E5%AE%9E%E7%8E%B0/ <h1 id="llm响应评估与spring-ai使用递归顾问构建llm-as-a-judge">LLM响应评估与Spring AI：使用递归顾问构建LLM-as-a-Judge</h1> <p>评估大型语言模型（LLM）输出的挑战对于 notoriously 非确定性的AI应用至关重要，特别是当它们进入生产环境时。像ROUGE和BLEU这样的传统指标在评估现代LLM产生的细致入微、上下文相关的响应时显得不足。人工评估虽然准确，但成本高、速度慢且无法扩展。</p>