# Dissecting Query-Key Interaction in Vision Transformers

arXiv: https://arxiv.org/abs/2405.14880 · HTML: https://mlsanity.com/paper/2405.14880/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 46 references that are in the map.
- First posted on arXiv 4 Apr 2024.

## Builds on (most cited first, 25 of 25)

- [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)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [SimMIM: A Simple Framework for Masked Image Modeling](https://mlsanity.com/paper/2111.09886/) (2021)
- [Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture](https://mlsanity.com/paper/2301.08243/) (2023)
- [How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddings](https://mlsanity.com/paper/1909.00512/) (2019)
- [Mind the Gap: Understanding the Modality Gap in Multi-modal Contrastive Representation Learning](https://mlsanity.com/paper/2203.02053/) (2022)
- [Deep ViT Features as Dense Visual Descriptors](https://mlsanity.com/paper/2112.05814/) (2021)
- [When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations](https://mlsanity.com/paper/2106.01548/) (2021)
- [Localizing Objects with Self-Supervised Transformers and no Labels](https://mlsanity.com/paper/2109.14279/) (2021)
- [Self-Supervised Transformers for Unsupervised Object Discovery using Normalized Cut](https://mlsanity.com/paper/2202.11539/) (2022)
- [Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla](https://mlsanity.com/paper/2307.09458/) (2023)
- [Analyzing Transformers in Embedding Space](https://mlsanity.com/paper/2209.02535/) (2022)
- [Revealing the Dark Secrets of Masked Image Modeling](https://mlsanity.com/paper/2205.13543/) (2022)
- [What do Vision Transformers Learn? A Visual Exploration](https://mlsanity.com/paper/2212.06727/) (2022)
- [IsoScore: Measuring the Uniformity of Embedding Space Utilization](https://mlsanity.com/paper/2108.07344/) (2021)
- [AttentionViz: A Global View of Transformer Attention](https://mlsanity.com/paper/2305.03210/) (2023)
- [Self-attention in Vision Transformers Performs Perceptual Grouping, Not Attention](https://mlsanity.com/paper/2303.01542/) (2023)
- [Do Saliency Models Detect Odd-One-Out Targets? New Datasets and Evaluations](https://mlsanity.com/paper/2005.06583/) (2020)
- [Affinity-based Attention in Self-supervised Transformers Predicts Dynamics of Object Grouping in Humans](https://mlsanity.com/paper/2306.00294/) (2023)
- [Is Anisotropy Inherent to Transformers?](https://mlsanity.com/paper/2306.07656/) (2023)

## Built on by (most cited first, 12 of 12)

- [Towards Interpreting Visual Information Processing in Vision-Language Models](https://mlsanity.com/paper/2410.07149/) (2024)
- [Bi-Orthogonal Factor Decomposition for Vision Transformers](https://mlsanity.com/paper/2601.05328/) (2026)
- [Decomposing Query-Key Feature Interactions Using Contrastive Covariances](https://mlsanity.com/paper/2602.04752/) (2026)
- [Causal Attribution via Activation Patching](https://mlsanity.com/paper/2603.13652/) (2026)
- [Mechanism-Driven Monitors for Preemptive Detection of LLM Training Instability](https://mlsanity.com/paper/2606.28116/) (2026)
- [Singular Vectors of Attention Heads Align with Features](https://mlsanity.com/paper/2602.13524/) (2026)
- [Diffusion-CAM: Faithful Visual Explanations for dMLLMs](https://mlsanity.com/paper/2604.11005/) (2026)
- [Intrinsic Interaction Geometry Controls the Low-Rank Complexity of Softmax Attention](https://mlsanity.com/paper/2608.28150/) (2026)
- [A Unified Geometric Space for Topological Alignment Between Transformer-Based Models and Human Brain Networks](https://mlsanity.com/paper/2510.24342/) (2025)
- [Enhancing Continual Learning of Vision-Language Models via Dynamic Prefix Weighting](https://mlsanity.com/paper/2604.18075/) (2026)
- [Prescriptive SVD-Inspired Attention via Spectral Energy Retention](https://mlsanity.com/paper/2609.24370/) (2026)
- [Query Expansion and Key Specialization in Transformer Attention Geometry](https://mlsanity.com/paper/2609.34273/) (2026)

## Abstract (arXiv)

Self-attention in vision transformers is often thought to perform perceptual grouping where tokens attend to other tokens with similar embeddings, which could correspond to semantically similar features of an object. However, attending to dissimilar tokens can be beneficial by providing contextual information. We propose to analyze the query-key interaction by the singular value decomposition of the interaction matrix (i.e. Wq^topWk). We find that in many ViTs, especially those with classification training objectives, early layers attend more to similar tokens, while late layers show increased attention to dissimilar tokens, providing evidence corresponding to perceptual grouping and contextualization, respectively. Many of these interactions between features represented by singular vectors are interpretable and semantic, such as attention between relevant objects, between parts of an object, or between the foreground and background. This offers a novel perspective on interpreting the attention mechanism, which contributes to understanding how transformer models utilize context and salient features when processing images.