On Vanishing Variance in Transformer Length Generalization

• URL: http://arxiv.org/abs/2504.02827v1
• Date: Thu, 03 Apr 2025 17:59:56 GMT
• Title: On Vanishing Variance in Transformer Length Generalization
• Authors: Ruining Li, Gabrijel Boduljak, Jensen, Zhou,
• Abstract summary: We show 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.<n>Our analyses attribute this improvement to a reduction-though not a complete elimination-of the distribution shift caused by vanishing variance.
• Score: 23.706900145711913
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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.

Related papers

• The Role of Sparsity for Length Generalization in Transformers [58.65997625433689]
  We propose a new theoretical framework to study length generalization for the next-token prediction task. We show that length generalization occurs as long as each predicted token depends on a small (fixed) number of previous tokens. We introduce Predictive Position Coupling, which trains the transformer to predict the position IDs used in a positional coupling approach.
  arXiv  Detail & Related papers  (2025-02-24T03:01:03Z)
• Differential Transformer [99.5117269150629]
  Transformer tends to overallocate attention to irrelevant context. We introduce Diff Transformer, which amplifies attention to relevant context while canceling noise. It offers notable advantages in practical applications, such as long-context modeling, key information retrieval, hallucination mitigation, in-context learning, and reduction of activation outliers.
  arXiv  Detail & Related papers  (2024-10-07T17:57:38Z)
• Transformers Can Achieve Length Generalization But Not Robustly [76.06308648699357]
  We show that the success of length generalization is intricately linked to the data format and the type of position encoding. We show for the first time that standard Transformers can extrapolate to a sequence length that is 2.5x the input length.
  arXiv  Detail & Related papers  (2024-02-14T18:18:29Z)
• From Interpolation to Extrapolation: Complete Length Generalization for Arithmetic Transformers [7.011373967209572]
  We show that transformer models are able to generalize to long lengths with the help of targeted attention biasing. We demonstrate that using ABC, the transformer model can achieve unprecedented near-perfect length generalization on certain arithmetic tasks.
  arXiv  Detail & Related papers  (2023-10-18T14:10:47Z)
• iTransformer: Inverted Transformers Are Effective for Time Series Forecasting [62.40166958002558]
  We propose iTransformer, which simply applies the attention and feed-forward network on the inverted dimensions. The iTransformer model achieves state-of-the-art on challenging real-world datasets.
  arXiv  Detail & Related papers  (2023-10-10T13:44:09Z)
• Robust representations of oil wells' intervals via sparse attention mechanism [2.604557228169423]
  We introduce the class of efficient Transformers named Regularized Transformers (Reguformers) The focus in our experiments is on oil&gas data, namely, well logs. To evaluate our models for such problems, we work with an industry-scale open dataset consisting of well logs of more than 20 wells.
  arXiv  Detail & Related papers  (2022-12-29T09:56:33Z)
• A Length-Extrapolatable Transformer [98.54835576985664]
  We focus on length extrapolation, i.e., training on short texts while evaluating longer sequences. We introduce a relative position embedding to explicitly maximize attention resolution. We evaluate different Transformer variants with language modeling.
  arXiv  Detail & Related papers  (2022-12-20T18:56:20Z)
• Sequence Length is a Domain: Length-based Overfitting in Transformer Models [0.0]
  In machine translation, the neural-based systems perform worse on very long sequences when compared to the preceding phrase-based translation approaches. We show that the observed drop in performance is due to the hypothesis length corresponding to the lengths seen by the model during training rather than the length of the input sequence.
  arXiv  Detail & Related papers  (2021-09-15T13:25:19Z)
• Finetuning Pretrained Transformers into RNNs [81.72974646901136]
  Transformers have outperformed recurrent neural networks (RNNs) in natural language generation. A linear-complexity recurrent variant has proven well suited for autoregressive generation. This work aims to convert a pretrained transformer into its efficient recurrent counterpart.
  arXiv  Detail & Related papers  (2021-03-24T10:50:43Z)
• Normalized Attention Without Probability Cage [12.18340575383456]
  We show limitations of constraining attention weights to the probability simplex. We propose to replace the softmax in self-attention with normalization. We support our insights with empirical results from more than 25,000 trained models.
  arXiv  Detail & Related papers  (2020-05-19T16:26:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.