Unverified50% confidenceFactExact time
深度神经网络天然倾向于过拟合训练分布，对分布外样本的处理能力往往大幅下降
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
The Truth Behind AI Benchmark Leaderboards: Real-World Performance Is the True Moat
bilibili卢菁博士_北大AI博士后7/7/2026
Related Claims
UnverifiedNeural networks have an overconfidence defect: even when inputs fall outside the training distribution, the Softmax layer may still output inflated confidence scores.81% similarUnverifiedLi等人2018年发现神经网络的有效参数空间维度远低于名义参数量，训练轨迹被限制在低维子空间中80% similarUnverifiedVanishing gradient refers to the exponential decay of gradient values during backpropagation in deep neural networks, causing parameters near the input layer to receive almost no updates.79% similarUnverified一段数据参与神经网络训练后，其影响以分布式方式弥散在数十亿参数中，难以被真正删除77% similarUnverifiedDuring cross-domain fine-tuning, an excessively large learning rate will destroy pre-trained weights, while an excessively small one will result in almost no learning, requiring a careful trade-off.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/428962API
curl https://kongchang.com/api/v1/knowledge/claims/428962MCP
get_claim(id=428962)