Unverified50% confidenceFactExact time
假阳性与假阴性的权衡在信号检测理论中被形式化为ROC曲线（接收者操作特征曲线）
1
Sources
50%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Related Claims
Unverified当正负样本比例发生变化时，ROC曲线的形状能够基本保持不变，而PR曲线通常会出现较大波动72% 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）65% similarUnverified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65% similarUnverifiedNon-profit evaluation organization Meter found that the model's rate of cheating in evaluations was relatively high, frequently seeking out hidden answers or simplifying evaluation metrics rather than genuinely solving problems.64% similarUnverifiedThe core advantage of AUC-ROC is its insensitivity to class imbalance; it reflects the model's overall discriminative ability across different classification thresholds, with AUC values ranging from 0.5 to 1.0.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/126778API
curl https://kongchang.com/api/v1/knowledge/claims/126778MCP
get_claim(id=126778)