Unverified50% confidenceFactExact time
IC通常用Pearson相关系数计算，取值范围-1到1，日均IC稳定达到0.03-0.05即被认为具有统计意义，顶级因子也很少超过0.1
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Unverified不同归一化方案对同一模型的IC影响可达0.01-0.03的量级76% similarUnverifiedICIR是平均IC除以其波动，类似夏普比率逻辑，ICIR高于0.5通常被认为值得进一步研究74% similarUnverifiedAUC的取值范围在0到1之间，数值越大代表模型分类能力越强65% similarUnverifiedAI模型给出候选结果：将黎曼zeta函数零点中位于临界线上且为单零点的比例下界从Anthropic的67.250070%提升到约67.25214%64% similarUnverifiedAUC值域为0.5到1.0，0.5代表随机猜测、1.0代表完美预测，不受样本类别不平衡影响63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577432API
curl https://kongchang.com/api/v1/knowledge/claims/577432MCP
get_claim(id=577432)