Unverified50% confidenceCautionExact time
The experiment has a small sample size (single model, single test, 50 questions), is subject to random fluctuation, and constitutes a heuristic finding rather than a definitive conclusion.
1
Sources
50%
Confidence
Long-term
Relevance
9/27/2026
First Seen
Sources
Related Entities
Related Claims
Unverified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79% similarUnverified在5%显著性水平下测试100个随机因子，期望会有5个看起来显著（多重假设检验问题）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75% similarUnverifiedWhen a team tests 1,000 combinations of strategy parameters, even if all combinations have zero predictive power, roughly 50 combinations will appear to perform excellently in backtesting purely due to randomness (the multiple comparisons problem).75% similarUnverifiedAnthropic的研究显示，即使面对毫无根据的质疑，在某些测试中模型被用户简单反驳后放弃正确答案的比例超过50%73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/954332API
curl https://kongchang.com/api/v1/knowledge/claims/954332MCP
get_claim(id=954332)