arXiv Machine Learning By Yize Zhao, Isabel Papadimitriou, Christos Thrampoulidis

Structure Before Collapse: Transient semantic geometry in next-token prediction

Read the original on arXiv Machine Learning →

arXiv:2606. 26749v1 Announce Type: new Abstract: Neural Collapse predicts that balanced one-hot classification pushes model representations to be equally far from each other; a symmetric configuration that depends only on the output label and ignores any semantic similarity in the inputs.

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

Hugging Face Trending Papers
Jun 25

Structure Before Collapse: Transient semantic geometry in next-token prediction

Neural Collapse predicts that balanced one-hot classification pushes model representations to be equally far from each other; a symmetric configuration that depends only on the output label and ignores any semantic similarity in the inputs. This creates a puzzle: next-token prediction language models are trained predominantly (as context length increases) with one-hot labels: the same context is very unlikely to appear twice in training with different labels.