Unverified50% confidenceCautionExact time
若所有陪审员来自同一模型家族、使用相同训练数据，它们很可能犯下相同错误，投票失去纠错意义
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
Unverified对同一测试用例执行多次调用统计成功比例得到通过率，比二元的通过/失败判断更能反映模型真实表现58% similarUnverified研究人员发现部分幻觉包名在10%以上的测试中反复出现，同一模型在相似提示下往往产生相同幻觉58% similarUnverifiedMcNemar 检验关注两个模型在每个样本上的分歧情况，只有判断结果不一致的样本（b和c单元格）具有信息量，两者都对或都错的样本被视为无关信息56% similarUnverified在事实性评估任务上模型评判者准确率仍显著低于人类专家，但在风格、连贯性等主观维度已接近人类评审员间一致性水平55% similarUnverifiedIn scenarios with extreme class imbalance (such as fraud detection), looking solely at accuracy as an evaluation metric is completely meaningless.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/513817API
curl https://kongchang.com/api/v1/knowledge/claims/513817MCP
get_claim(id=513817)