Incremental Learning of Sparse Attention Patterns in Transformers

• URL: http://arxiv.org/abs/2602.19143v1
• Date: Sun, 22 Feb 2026 12:16:06 GMT
• Title: Incremental Learning of Sparse Attention Patterns in Transformers
• Authors: Oğuz Kaan Yüksel, Rodrigo Alvarez Lucendo, Nicolas Flammarion,
• Abstract summary: This paper introduces a high-order Markov chain task to investigate how transformers learn to integrate information from multiple past positions.<n>We identify a shift in learning dynamics from competitive, where heads converge on the most statistically dominant pattern, to cooperative, where heads specialize in distinct patterns.
• Score: 29.54151079577767
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: This paper introduces a high-order Markov chain task to investigate how transformers learn to integrate information from multiple past positions with varying statistical significance. We demonstrate that transformers learn this task incrementally: each stage is defined by the acquisition of specific information through sparse attention patterns. Notably, we identify a shift in learning dynamics from competitive, where heads converge on the most statistically dominant pattern, to cooperative, where heads specialize in distinct patterns. We model these dynamics using simplified differential equations that characterize the trajectory and prove stage-wise convergence results. Our analysis reveals that transformers ascend a complexity ladder by passing through simpler, misspecified hypothesis classes before reaching the full model class. We further show that early stopping acts as an implicit regularizer, biasing the model toward these simpler classes. These results provide a theoretical foundation for the emergence of staged learning and complex behaviors in transformers, offering insights into generalization for natural language processing and algorithmic reasoning.

Related papers

• The Kinetics of Reasoning: How Chain-of-Thought Shapes Learning in Transformers? [25.29458951592086]
  Chain-of-thought (CoT) supervision can substantially improve transformer performance.<n>We investigate these learning dynamics through the lens of grokking by pretraining transformers on symbolic reasoning tasks.
  arXiv  Detail & Related papers  (2025-10-28T20:14:26Z)
• Transformers as Multi-task Learners: Decoupling Features in Hidden Markov Models [12.112842686827669]
  Transformer based models have shown remarkable capabilities in sequence learning across a wide range of tasks.<n>We investigate the layerwise behavior of Transformers to uncover the mechanisms underlying their multi-task generalization ability.<n>Our explicit constructions align closely with empirical observations, providing theoretical support for the Transformer's effectiveness and efficiency on sequence learning across diverse tasks.
  arXiv  Detail & Related papers  (2025-06-02T17:39:31Z)
• A distributional simplicity bias in the learning dynamics of transformers [50.91742043564049]
  We show that transformers, trained on natural language data, also display a simplicity bias.<n>Specifically, they sequentially learn many-body interactions among input tokens, reaching a saturation point in the prediction error for low-degree interactions.<n>This approach opens up the possibilities of studying how interactions of different orders in the data affect learning, in natural language processing and beyond.
  arXiv  Detail & Related papers  (2024-10-25T15:39:34Z)
• Interpreting Affine Recurrence Learning in GPT-style Transformers [54.01174470722201]
  In-context learning allows GPT-style transformers to generalize during inference without modifying their weights. This paper focuses specifically on their ability to learn and predict affine recurrences as an ICL task. We analyze the model's internal operations using both empirical and theoretical approaches.
  arXiv  Detail & Related papers  (2024-10-22T21:30:01Z)
• In-Context Learning with Representations: Contextual Generalization of Trained Transformers [66.78052387054593]
  In-context learning (ICL) refers to a capability of pretrained large language models, which can learn a new task given a few examples during inference. This paper investigates the training dynamics of transformers by gradient descent through the lens of non-linear regression tasks.
  arXiv  Detail & Related papers  (2024-08-19T16:47:46Z)
• Learning on Transformers is Provable Low-Rank and Sparse: A One-layer Analysis [63.66763657191476]
  We show that efficient numerical training and inference algorithms as low-rank computation have impressive performance for learning Transformer-based adaption. We analyze how magnitude-based models affect generalization while improving adaption. We conclude that proper magnitude-based has a slight on the testing performance.
  arXiv  Detail & Related papers  (2024-06-24T23:00:58Z)
• Local to Global: Learning Dynamics and Effect of Initialization for Transformers [20.02103237675619]
  We focus on first-order Markov chains and single-layer transformers. We prove that transformer parameters trained on next-token prediction loss can either converge to global or local minima.
  arXiv  Detail & Related papers  (2024-06-05T08:57:41Z)
• In-Context Convergence of Transformers [63.04956160537308]
  We study the learning dynamics of a one-layer transformer with softmax attention trained via gradient descent. For data with imbalanced features, we show that the learning dynamics take a stage-wise convergence process.
  arXiv  Detail & Related papers  (2023-10-08T17:55:33Z)
• Topographic VAEs learn Equivariant Capsules [84.33745072274942]
  We introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to organize its activations according to salient characteristics such as digit class, width, and style on MNIST. We demonstrate approximate equivariance to complex transformations, expanding upon the capabilities of existing group equivariant neural networks.
  arXiv  Detail & Related papers  (2021-09-03T09:25:57Z)
• Masked Language Modeling for Proteins via Linearly Scalable Long-Context Transformers [42.93754828584075]
  We present a new Transformer architecture, Performer, based on Fast Attention Via Orthogonal Random features (FAVOR) Our mechanism scales linearly rather than quadratically in the number of tokens in the sequence, is characterized by sub-quadratic space complexity and does not incorporate any sparsity pattern priors. It provides strong theoretical guarantees: unbiased estimation of the attention matrix and uniform convergence.
  arXiv  Detail & Related papers  (2020-06-05T17:09:16Z)

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.