# The Cow of Rembrandt - Analyzing Artistic Prompt Interpretation in Text-to-Image Models

arXiv: https://arxiv.org/abs/2507.23313 · HTML: https://mlsanity.com/paper/2507.23313/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models.
- Builds on 19 of its 24 references that are in the map.
- First posted on arXiv 31 Jul 2025.

## Placed in

- Diffusion models / [Image generation, text-to-image and editing](https://mlsanity.com/diffusion/image-generation-text-to-image-and-editing/)

## Builds on (most cited first, 19 of 19)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [What the DAAM: Interpreting Stable Diffusion Using Cross Attention](https://mlsanity.com/paper/2210.04885/) (2022)
- [Design Guidelines for Prompt Engineering Text-to-Image Generative Models](https://mlsanity.com/paper/2109.06977/) (2021)
- [Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Feature](https://mlsanity.com/paper/1505.00855/) (2015)
- [Diffuse, Attend, and Segment: Unsupervised Zero-Shot Segmentation using Stable Diffusion](https://mlsanity.com/paper/2308.12469/) (2023)
- [Measuring Style Similarity in Diffusion Models](https://mlsanity.com/paper/2404.01292/) (2024)
- [Diffusion Model is Secretly a Training-free Open Vocabulary Semantic Segmenter](https://mlsanity.com/paper/2309.02773/) (2023)
- [A Taxonomy of Prompt Modifiers for Text-To-Image Generation](https://mlsanity.com/paper/2204.13988/) (2022)
- [DiffusionSeg: Adapting Diffusion Towards Unsupervised Object Discovery](https://mlsanity.com/paper/2303.09813/) (2023)
- [ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features](https://mlsanity.com/paper/2502.04320/) (2025)
- [Style and Content Disentanglement in Generative Adversarial Networks](https://mlsanity.com/paper/1811.05621/) (2018)
- [Not Only Generative Art: Stable Diffusion for Content-Style Disentanglement in Art Analysis](https://mlsanity.com/paper/2304.10278/) (2023)
- [Improving Style-Content Disentanglement in Image-to-Image Translation](https://mlsanity.com/paper/2007.04964/) (2020)
- [Identifying and Analyzing Performance-Critical Tokens in Large Language Models](https://mlsanity.com/paper/2401.11323/) (2024)

## Built on by (most cited first, 2 of 2)

- [LumiXAI: A Modular Full-Stack Framework for Feature Attribution](https://mlsanity.com/paper/2608.24524/) (2026)
- [Fragment-Aware Vision Transformers for Fresco-Fragment Style Classification](https://mlsanity.com/paper/2609.21012/) (2026)

## Abstract (arXiv)

Text-to-image diffusion models have demonstrated remarkable capabilities in generating artistic content by learning from billions of images, including popular artworks. However, the fundamental question of how these models internally represent concepts, such as content and style in paintings, remains unexplored. Traditional computer vision assumes content and style are orthogonal, but diffusion models receive no explicit guidance about this distinction during training. In this work, we investigate how transformer-based text-to-image diffusion models encode content and style concepts when generating artworks. We leverage cross-attention heatmaps to attribute pixels in generated images to specific prompt tokens, enabling us to isolate image regions influenced by content-describing versus style-describing tokens. Our findings reveal that diffusion models demonstrate varying degrees of content-style separation depending on the specific artistic prompt and style requested. In many cases, content tokens primarily influence object-related regions while style tokens affect background and texture areas, suggesting an emergent understanding of the content-style distinction. These insights contribute to our understanding of how large-scale generative models internally represent complex artistic concepts without explicit supervision. We share the code and dataset, together with an exploratory tool for visualizing attention maps at https://github.com/umilISLab/artistic-prompt-interpretation.