Unverified50% confidenceFactExact time
The theoretical assumption of traditional supervised learning is that training data and test data come from the same distribution (the i.i.d. assumption), but this almost never holds in reality.
1
Sources
50%
Confidence
Long-term
Relevance
9/1/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedWhen a model's training data contains benchmark test questions or highly similar content, the model may achieve high scores through memorization rather than genuine understanding.72% similarUnverified模型本身对准确性毫无概念，它只是模仿训练数据中观察到的统计分布71% similarUnverified标准的 McNemar 检验假设比较的是两个确定性模型在单一固定测试集上的预测结果，而随机种子引入的是模型训练本身的变异性，这两种不确定性来源在概念上属于不同层次69% similarUnverified蒸馏训练范式导致开源模型学习到经过高度过滤、几乎不包含失败和纠错过程的数据分布，使模型形成不愿自我纠正的性格68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/838966API
curl https://kongchang.com/api/v1/knowledge/claims/838966MCP
get_claim(id=838966)