Unverified50% confidenceBenchmarkExact time
k=5时可以以95%置信度区分成功率0.9和0.7的模型，k=10时能识别更细微的差异
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
QuantaMind：本地验证AI Agent可靠性的开源测试工具
Related Claims
UnverifiedMinimizing cross-entropy is equivalent to minimizing the KL divergence between the true distribution and the predicted distribution.63% similarUnverified报告同时给出中位数KLD和99.9%分位KLD两个数值，前者反映典型情况，后者捕捉极端长尾错误63% similarUnverifiedLLM每次调用可建模为伯努利试验，N次调用通过次数服从二项分布B(N, p)，通过率95%置信区间约为p̂ ± 1.96×√(p̂(1-p̂)/N)61% similarUnverified在量化内容分析中，编码者间信度通常使用Cohen's Kappa或Krippendorff's Alpha系数衡量，一般要求达到0.7以上才算可接受60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/510666API
curl https://kongchang.com/api/v1/knowledge/claims/510666MCP
get_claim(id=510666)