check-answer-consistency

Use this to get a cheap, reference-free signal that an LLM answer might be made up, by sampling the same prompt a few times and measuring agreement. Trigger on "is this answer reliable", "flag low-confidence answers", "cheap hallucination check", "confidence score without a ground truth", "self-consistency check". Ships a runnable, tested scorer you can put inline or on sampled traffic.

适合你，如果需要在无标准答案时快速判断LLM输出可信度