Unverified50% confidenceFactExact time
Bonferroni 校正是最保守的多重比较校正方案，将显著性阈值除以比较次数；Benjamini-Hochberg FDR 校正更宽松，控制所有显著结果中假阳性的比例
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
Unverified统计功效（Statistical Power）通常设为 80% 以上，多重比较问题可通过 Bonferroni 校正或 FDR 控制等方法调整71% similarUnverifiedThe 2023 TruthfulQA benchmark showed that even the strongest models at the time did not exceed 60% accuracy on questions requiring precise factual judgment.62% similarUnverified泰特洛克的研究得出结论：专家预测的准确率仅略高于随机猜测，甚至不如简单的统计基准线60% similarUnverified探索性能上界实验发现现有验证粒度并非最优，理想情况下计算量（FLOPs）可减少75倍，同时准确率提升1.1%59% similarUnverifiedIn class-imbalance scenarios, a model can achieve seemingly impressive accuracy by always predicting the majority class, a phenomenon known as the Accuracy Paradox.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/502213API
curl https://kongchang.com/api/v1/knowledge/claims/502213MCP
get_claim(id=502213)