Saved in:
Bibliographic Details
Main Authors: Li, Ruining, Boduljak, Gabrijel, Jensen, Zhou
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2504.02827
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916673318027264
author Li, Ruining
Boduljak, Gabrijel
Jensen
Zhou
author_facet Li, Ruining
Boduljak, Gabrijel
Jensen
Zhou
contents It is a widely known issue that Transformers, when trained on shorter sequences, fail to generalize robustly to longer ones at test time. This raises the question of whether Transformer models are real reasoning engines, despite their impressive abilities in mathematical problem solving and code synthesis. In this paper, we offer a vanishing variance perspective on this issue. To the best of our knowledge, we are the first to demonstrate that even for today's frontier models, a longer sequence length results in a decrease in variance in the output of the multi-head attention modules. On the argmax retrieval and dictionary lookup tasks, our experiments show that applying layer normalization after the attention outputs leads to significantly better length generalization. Our analyses attribute this improvement to a reduction-though not a complete elimination-of the distribution shift caused by vanishing variance.
format Preprint
id arxiv_https___arxiv_org_abs_2504_02827
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On Vanishing Variance in Transformer Length Generalization
Li, Ruining
Boduljak, Gabrijel
Jensen
Zhou
Machine Learning
Artificial Intelligence
It is a widely known issue that Transformers, when trained on shorter sequences, fail to generalize robustly to longer ones at test time. This raises the question of whether Transformer models are real reasoning engines, despite their impressive abilities in mathematical problem solving and code synthesis. In this paper, we offer a vanishing variance perspective on this issue. To the best of our knowledge, we are the first to demonstrate that even for today's frontier models, a longer sequence length results in a decrease in variance in the output of the multi-head attention modules. On the argmax retrieval and dictionary lookup tasks, our experiments show that applying layer normalization after the attention outputs leads to significantly better length generalization. Our analyses attribute this improvement to a reduction-though not a complete elimination-of the distribution shift caused by vanishing variance.
title On Vanishing Variance in Transformer Length Generalization
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2504.02827