Unverified50% confidenceFactExact time
叠加现象指神经网络倾向于用远少于特征数量的神经元同时编码大量语义特征，与传统一个神经元一个特征的直觉相悖
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
因果理论破解大模型黑箱：机械可解释性的新范式
Related Claims
Unverified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72% similarUnverified蒸馏的核心洞察在于神经网络的软标签（类别概率分布）包含比one-hot硬标签更丰富的结构信息68% similarUnverified神经网络中各层对精度损失的敏感程度差异显著，patch projection等输入投影层敏感度高，中间主干的注意力与FFN模块对低精度更鲁棒67% similarUnverified线性代数中的矩阵分解与特征值分解对应神经网络的权重空间分析67% similarUnverifiedAn intermittently faulty memory module may alternately produce correct and incorrect computation results, causing sporadic deviations in gradient calculations during neural network training, affecting model convergence without triggering hardware alerts.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/509900API
curl https://kongchang.com/api/v1/knowledge/claims/509900MCP
get_claim(id=509900)