Unverified50% confidenceDecision RuleExact time
评估基准测试质量的关键指标是假阳性率和假阴性率,理想基准应将两者控制在较低水平
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
Related Claims
UnverifiedIn scenarios with extreme class imbalance (such as fraud detection), looking solely at accuracy as an evaluation metric is completely meaningless.73% similarUnverified假阳性(False Positive)指模型将负类样本(正常内容)错误预测为正类(违规内容)73% similarUnverified困惑度是测试集上每个token平均负对数似然的指数,值越低代表模型对语料的预测越准确71% similarUnverified最优阈值可通过最大化Youden指数(灵敏度+特异度-1)确定,在假阳性与假阴性代价对等时是最优阈值选择准则71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/704187API
curl https://kongchang.com/api/v1/knowledge/claims/704187MCP
get_claim(id=704187)