Finding Features in Neural Networks with the Empirical NTK
"Grokking" 相关文章
$\mathbf{Li_2}$: A Framework on Dynamics of Feature Emergence and Delayed Generalization
黑客滥用 X 的 AI 助手传播恶意链接
Tracing the Path to Grokking: Embeddings, Dropout, and Network Activation
Muon Optimizer Significantly Accelerates Grokking in Transformers: Microsoft Researchers Explore Optimizer Influence on Delayed Generalization
This AI Research from Ohio State University and CMU Discusses Implicit Reasoning in Transformers And Achieving Generalization Through Grokking
Grokfast：通过放大慢梯度加速格罗克学习
