Skip to content
AI Atlas
PaperActive

Output Embedding Centering for Stable LLM Pretraining

arxiv.org/abs/2601.02031

quality89

Updated 2 h ago · first seen 11 Sept 2026

paper_01M294FRY089Z3Z76WVA5WKM9N

Published
11 Sept 2026
T1 · 2 h ago
arXiv
2601.02031
T1 · 2 h ago
Category
cs.LG
T1 · 2 h ago

Abstract

Pretraining of large language models is not only expensive but also prone to certain training instabilities. A specific instability that often occurs at the end of training is output logit divergence. The most widely used mitigation strategies, z-loss and logit soft-capping, merely address the symptoms rather than the underlying cause of the problem. In this paper, we analyze the instability from the perspective of the output embeddings' geometry and identify anisotropic embeddings as its source. Based on this, we propose output embedding centering (OEC) as a new mitigation strategy, and demonstrate that it suppresses output logit divergence. OEC can be implemented in two different ways: as a deterministic operation called $\mu$-centering, or a regularization method called $\mu$-loss. Our experiments show that both variants outperform z-loss in terms of training stability, while being on par with logit soft-capping. This holds true both in the presence and the absence of weight tying. As a secondary result, we find that $\mu$-loss is significantly less sensitive to regularization hyperparameter tuning than z-loss.

Authors 3

Felix Stollenwerk, Anna Lokrantz, Niclas Hertzberg

Specification

Official page

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

Arxiv announce type
replace

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

arXiv id
2601.02031

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

Categories
cs.LG, cs.AI, cs.CL

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

PDF

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

Primary category
cs.LG

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

Published
11 Sept 2026

Source:arXiv (Atom API + RSS)T1observed 2 h agohigh

Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →

Provenance

Attributed facts

9

Source tiers

T19

Freshest observation

2 h ago

Conflicts

None