Unverified50% confidenceSolutionExact time
在机器学习多种子实验中重复次数通常只有5到10次，样本量极小，正态性假设难以验证，因此 Wilcoxon 检验往往是更保守稳健的选择
1
Sources
50%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
McNemar检验对比机器学习模型：随机种子下的正确用法
redditr/learnmachinelearning7/11/2026
Related Claims
UnverifiedAI测试用例不能只执行一次就判定，业界通常建议同一测试用例至少执行5-10次以统计输出的稳定性和质量分布76% similarUnverified置换检验通过大量随机置换（通常1000到10000次）构建零假设下的经验分布来估计p值，无需参数假设但计算成本相对较高73% similarUnverifiedMultiple independent studies in 2023 showed that some top models scored extremely high on standard mathematical reasoning tests but performed significantly worse on equivalent rephrased versions, indicating that the models learned test patterns rather than genuine reasoning ability.71% similarUnverified该测试样本规模相对有限（5个模型、13个代码库），结论有待更大规模验证71% similarUnverified多重比较问题指同时测试多个模型变体时，即使每个单独检验显著性水平设为0.05，整体假阳性率也会急剧膨胀71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/502212API
curl https://kongchang.com/api/v1/knowledge/claims/502212MCP
get_claim(id=502212)