Unverified50% confidenceFactExact time
When the label error rate is below about 10%-20%, deep networks can still learn effective representations, and the degradation in generalization performance is limited.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn an S&P 500 option chain experiment, after sparsifying the data to retain only 20 quotes and introducing no-arbitrage constraints, the out-of-sample extrapolation error on the held-out set decreased by 22% and the calendar arbitrage occurrence rate dropped from 10% to 1%.72% similarUnverified残差连接支撑了152层超深网络的训练，并在2015年ImageNet竞赛中以低于4%的错误率首次超越人类水平（约5%）70% similarUnverifiedTop academic deepfake detectors can see their accuracy plummet from 99% to below 60% when faced with previously unseen generative models, a phenomenon known as the cross-model generalization problem.69% similarUnverified推测解码中当草稿通过率低于50%时，实际加速效果会大打折扣甚至产生负优化69% similarUnverified实验数据显示，经过5-10轮迭代训练后，模型在低频词汇和长尾知识上的覆盖率下降超过40%，而输出文本在表面流畅性上几乎没有可察觉的退化69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1038949API
curl https://kongchang.com/api/v1/knowledge/claims/1038949MCP
get_claim(id=1038949)