# Learning to Read the Contextual Tokens in Diffusion Transformers

arXiv: https://arxiv.org/abs/2610.06844 · HTML: https://mlsanity.com/paper/2610.06844/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models.
- Builds on 26 of its 72 references that are in the map.
- First posted on arXiv 5 Oct 2026.

## 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, 26 of 26)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Hierarchical Text-Conditional Image Generation with CLIP Latents](https://mlsanity.com/paper/2204.06125/) (2022)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Prompt-to-Prompt Image Editing with Cross Attention Control](https://mlsanity.com/paper/2208.01626/) (2022)
- [SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations](https://mlsanity.com/paper/2108.01073/) (2021)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [HunyuanVideo: A Systematic Framework For Large Video Generative Models](https://mlsanity.com/paper/2412.03603/) (2024)
- [Demystifying MMD GANs](https://mlsanity.com/paper/1801.01401/) (2018)
- [Qwen-Image Technical Report](https://mlsanity.com/paper/2508.02324/) (2025)
- [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://mlsanity.com/paper/2506.15742/) (2025)
- [Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think](https://mlsanity.com/paper/2410.06940/) (2024)
- [Human Preference Score: Better Aligning Text-to-Image Models with Human Preference](https://mlsanity.com/paper/2303.14420/) (2023)
- [HiDream-I1: A High-Efficient Image Generative Foundation Model with Sparse Diffusion Transformer](https://mlsanity.com/paper/2505.22705/) (2025)
- [VideoJAM: Joint Appearance-Motion Representations for Enhanced Motion Generation in Video Models](https://mlsanity.com/paper/2502.02492/) (2025)
- [No Other Representation Component Is Needed: Diffusion Transformers Can Provide Representation Guidance by Themselves](https://mlsanity.com/paper/2505.02831/) (2025)
- [ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features](https://mlsanity.com/paper/2502.04320/) (2025)
- [HunyuanVideo-Foley: Multimodal Diffusion with Representation Alignment for High-Fidelity Foley Audio Generation](https://mlsanity.com/paper/2508.16930/) (2025)
- [Kling-Foley: Multimodal Diffusion Transformer for High-Quality Video-to-Audio Generation](https://mlsanity.com/paper/2506.19774/) (2025)
- [Qwen-Image-2.0 Technical Report](https://mlsanity.com/paper/2605.10730/) (2026)
- [Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis](https://mlsanity.com/paper/2603.06507/) (2026)
- [Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelines](https://mlsanity.com/paper/2403.05846/) (2024)
- [Fine-T2I: An Open, Large-Scale, and Diverse Dataset for High-Quality T2I Fine-Tuning](https://mlsanity.com/paper/2602.09439/) (2026)
- [Text Template Tokens Are Implicit Semantic Registers in Diffusion Transformers](https://mlsanity.com/paper/2607.19139/) (2026)

## Abstract (arXiv)

Multimodal Diffusion Transformers (MM-DiTs) jointly process visual and textual representations throughout generation. These models repeatedly update the text tokens through multimodal attention, forming dynamic contextual tokens whose function is not well understood. In this work, we introduce a framework for reading this contextual space through natural-language interrogation. We train a lightweight bottleneck network that maps intermediate contextual tokens into the input space of a frozen Large Language Model (LLM), allowing the LLM to answer questions about the emerging image directly from these hidden representations. Our reader reveals that contextual tokens encode a rich, global representation of the emerging scene: generation-specific semantics, including attributes left underspecified by the prompt, are accessible surprisingly early in denoising, while increasingly fine-grained details become readable over time. Remarkably, this information remains decodable even when the MM-DiT receives an empty prompt, showing that contextual tokens accumulate substantial image-specific information from the evolving visual representation itself. We further find that generations with more readable contextual representations tend to receive higher human-preference scores. Building on these observations, we introduce Contextual Alignment, a training technique that explicitly reinforces the visual-semantic information encoded in the contextual tokens, improving generation quality and distributional coverage. Together, our results establish contextual tokens as both an interpretable view into the internal dynamics of MM-DiTs and an effective target for improving generative models.