# Artifacts and Attention Sinks: Structured Approximations for Efficient Vision Transformers

Source: [Artifacts and Attention Sinks: Structured Approximations for Efficient Vision Transformers](<https://www.sota2.com/research/paper/68fa7a2e1e209131cea8198c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T05:14:13.950Z

Authors: Andrew Lu, Wentinn Liao, Liuhui Wang, Huzheng Yang, Jianbo Shi

Published: 2025-07-21T19:29:03.000Z

## Abstract

Vision transformers have emerged as a powerful tool across a wide range of applications, yet their inner workings remain only partially understood. In this work, we examine the phenomenon of massive tokens - tokens with exceptionally high activation norms that act as attention sinks - and artifact tokens that emerge as a byproduct during inference. Our analysis reveals that these tokens mutually suppress one another through the attention mechanism, playing a critical role in regulating information flow within the network. Leveraging these insights, we introduce Fast Nystr\\"om Attention (FNA), a training-free method that approximates self-attention in linear time and space by exploiting the structured patterns formed by massive and artifact tokens. Additionally, we propose a masking strategy to mitigate noise from these tokens, yielding modest performance gains at virtually no cost. We evaluate our approach on popular pretrained vision backbones and demonstrate competitive performance on retrieval, classification, segmentation, and visual question answering (VQA), all while reducing computational overhead.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2507.16018>)

[PDF](<https://arxiv.org/pdf/2507.16018.pdf>)
