Unverified50% confidenceFactExact time
召回率的计算公式为真正例除以真正例加假反例之和
1
Sources
50%
Confidence
Long-term
Relevance
8/29/2026
First Seen
Sources
Related Entities
Related Claims
Unverified在反向模式下，当模型参数数量高达数百万时，计算梯度的成本仅与一次前向计算相当65% similarUnverified测试时计算扩展的边际收益并非线性，存在明显的收益递减区间63% similarUnverifiedWhen large numbers of users call the same underlying model with similar prompts, convergence of outputs is almost a mathematical inevitability due to the consistent underlying weights and probability distributions.61% similarUnverifiedIn the no-show prediction task, the cost of missing truly absent patients (false negatives) is far higher than the cost of misclassifying patients who do show up, so recall should be prioritized.61% similarUnverified整体质量分数应趋势化和相对化展示（用百分比变化），而非展示绝对分数59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/817588API
curl https://kongchang.com/api/v1/knowledge/claims/817588MCP
get_claim(id=817588)