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

arXiv: https://arxiv.org/abs/2507.16018 · HTML: https://mlsanity.com/paper/2507.16018/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 32 references that are in the map.
- First posted on arXiv 21 Jul 2025.

## Builds on (most cited first, 20 of 20)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [VQA: Visual Question Answering](https://mlsanity.com/paper/1505.00468/) (2015)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [Vision Transformers Need Registers](https://mlsanity.com/paper/2309.16588/) (2023)
- [Nystr"omformer: A Nystr"om-Based Algorithm for Approximating Self-Attention](https://mlsanity.com/paper/2102.03902/) (2021)
- [Massive Activations in Large Language Models](https://mlsanity.com/paper/2402.17762/) (2024)
- [EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision](https://mlsanity.com/paper/2311.02077/) (2023)
- [When Attention Sink Emerges in Language Models: An Empirical View](https://mlsanity.com/paper/2410.10781/) (2024)
- [Learning A Sparse Transformer Network for Effective Image Deraining](https://mlsanity.com/paper/2303.11950/) (2023)
- [Combiner: Full Attention Transformer with Sparse Computation Cost](https://mlsanity.com/paper/2107.05768/) (2021)
- [The Super Weight in Large Language Models](https://mlsanity.com/paper/2411.07191/) (2024)
- [Denoising Vision Transformers](https://mlsanity.com/paper/2401.02957/) (2024)

## Built on by (most cited first, 10 of 10)

- [Attention Sinks Are Provably Necessary in Softmax Transformers: Evidence from Trigger-Conditional Tasks](https://mlsanity.com/paper/2603.11487/) (2026)
- [Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation](https://mlsanity.com/paper/2604.10098/) (2026)
- [Enhancing Concept Localization in CLIP-based Concept Bottleneck Models](https://mlsanity.com/paper/2510.07115/) (2025)
- [Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sink](https://mlsanity.com/paper/2602.05228/) (2026)
- [DAVE: Distribution-aware Attribution via ViT Gradient Decomposition](https://mlsanity.com/paper/2602.06613/) (2026)
- [When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models](https://mlsanity.com/paper/2604.03316/) (2026)
- [Activation Quantization of Vision Encoders Needs Prefixing Registers](https://mlsanity.com/paper/2510.04547/) (2025)
- [Order Matters: LVLMs as Judges for Temporal Reasoning in Image Sequences](https://mlsanity.com/paper/2608.10908/) (2026)
- [Sink-Token-Aware Pruning for Fine-Grained Video Understanding in Efficient Video LLMs](https://mlsanity.com/paper/2604.20937/) (2026)
- [A Unifying View of Attention Sinks: From Mechanisms to Architectural Interventions](https://mlsanity.com/paper/2606.08105/) (2026)

## Abstract (arXiv)

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.