arXiv Machine Learning By Xingjian Wang, Qingyu Han, Xiaodong Luo, Yin Zhang

Why Post-Norm Transformers Collapse: Attention Amplification and Gradient Repair Failure

Read the original on arXiv Machine Learning →

arXiv:2608. 09417v2 Announce Type: replace Abstract: Deep decoder-only Transformers often replace the original Post-Norm architecture with Pre-Norm variants because Post-Norm training is highly sensitive to warmup and learning rate under conventional initialization schemes.

Summary generated by The Flow from the publisher's feed. The full article lives at arXiv Machine Learning.

Hugging Face Trending Papers
Jul 15

Transforming Rank: How Architecture Navigates the Spectral Pathologies of Depth

We investigate how each component of the Transformer feedforward block architecture design determines how much rank survives across depth at initialization. We reinterpret skip connections and normalization, long understood as controlling magnitude, as mechanisms for preserving gradient rank across depth, since the very matrix multiplications and nonlinear activations that make the network expressive also reduce the rank.