# Efficient Diffusion Transformer with Step-wise Dynamic Attention Mediators

arXiv: https://arxiv.org/abs/2408.05710 · HTML: https://mlsanity.com/paper/2408.05710/

## Key facts

- Cited by 24 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 81 of its 92 references that are in the map.
- First posted on arXiv 11 Aug 2024.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)

## Builds on (most cited first, 30 of 81)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [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)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [Segment Anything](https://mlsanity.com/paper/2304.02643/) (2023)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [End-to-End Object Detection with Transformers](https://mlsanity.com/paper/2005.12872/) (2020)
- [Hierarchical Text-Conditional Image Generation with CLIP Latents](https://mlsanity.com/paper/2204.06125/) (2022)
- [Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding](https://mlsanity.com/paper/2205.11487/) (2022)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Zero-Shot Text-to-Image Generation](https://mlsanity.com/paper/2102.12092/) (2021)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Large Scale GAN Training for High Fidelity Natural Image Synthesis](https://mlsanity.com/paper/1809.11096/) (2018)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [Cascaded Diffusion Models for High Fidelity Image Generation](https://mlsanity.com/paper/2106.15282/) (2021)
- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis](https://mlsanity.com/paper/2310.00426/) (2023)
- [Are Sixteen Heads Really Better than One?](https://mlsanity.com/paper/1905.10650/) (2019)
- [SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers](https://mlsanity.com/paper/2401.08740/) (2024)
- [EVA: Exploring the Limits of Masked Visual Representation Learning at Scale](https://mlsanity.com/paper/2211.07636/) (2022)

## Built on by (most cited first, 24 of 24)

- [Training-free and Adaptive Sparse Attention for Efficient Long Video Generation](https://mlsanity.com/paper/2502.21079/) (2025)
- [Dynamic Diffusion Transformer](https://mlsanity.com/paper/2410.03456/) (2024)
- [ART: Anonymous Region Transformer for Variable Multi-Layer Transparent Image Generation](https://mlsanity.com/paper/2502.18364/) (2025)
- [Layer- and Timestep-Adaptive Differentiable Token Compression Ratios for Efficient Diffusion Transformers](https://mlsanity.com/paper/2412.16822/) (2024)
- [FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model](https://mlsanity.com/paper/2410.13925/) (2024)
- [DSV: Exploiting Dynamic Sparsity to Accelerate Large-Scale Video DiT Training](https://mlsanity.com/paper/2502.07590/) (2025)
- [Analysis of Attention in Video Diffusion Transformers](https://mlsanity.com/paper/2504.10317/) (2025)
- [SeaCache: Spectral-Evolution-Aware Cache for Accelerating Diffusion Models](https://mlsanity.com/paper/2602.18993/) (2026)
- [SparseDiT: Token Sparsification for Efficient Diffusion Transformer](https://mlsanity.com/paper/2412.06028/) (2024)
- [LiT: Delving into a Simple Linear Diffusion Transformer for Image Generation](https://mlsanity.com/paper/2501.12976/) (2025)
- [Qihoo-T2X: An Efficient Proxy-Tokenized Diffusion Transformer for Text-to-Any-Task](https://mlsanity.com/paper/2409.04005/) (2024)
- [ENAT: Rethinking Spatial-temporal Interactions in Token-based Image Synthesis](https://mlsanity.com/paper/2411.06959/) (2024)
- [DiCo: Revitalizing ConvNets for Scalable and Efficient Diffusion Modeling](https://mlsanity.com/paper/2505.11196/) (2025)
- [ProxyTransformation: Preshaping Point Cloud Manifold With Proxy Attention For 3D Visual Grounding](https://mlsanity.com/paper/2502.19247/) (2025)
- [Linear Differential Vision Transformer: Learning Visual Contrasts via Pairwise Differentials](https://mlsanity.com/paper/2511.00833/) (2025)
- [Norm×Direction: Restoring the Missing Query Norm in Vision Linear Attention](https://mlsanity.com/paper/2506.21137/) (2025)
- [TurboVSR: Fantastic Video Upscalers and Where to Find Them](https://mlsanity.com/paper/2506.23618/) (2025)
- [Towards Redundancy Reduction in Diffusion Models for Efficient Video Super-Resolution](https://mlsanity.com/paper/2509.23980/) (2025)
- [Designing Parameter and Compute Efficient Diffusion Transformers using Distillation](https://mlsanity.com/paper/2502.14226/) (2025)
- [Optimizing Inference in Transformer-Based Models: A Multi-Method Benchmark](https://mlsanity.com/paper/2509.17894/) (2025)
- [Dynamic Differential Linear Attention: Enhancing Linear Diffusion Transformer for High-Quality Image Generation](https://mlsanity.com/paper/2601.13683/) (2026)
- [LINA: Linear Autoregressive Image Generative Models with Continuous Tokens](https://mlsanity.com/paper/2601.22630/) (2026)
- [MirrorLA: Reflecting Feature Map for Vision Linear Attention](https://mlsanity.com/paper/2602.04346/) (2026)
- [Principal-timestep Restricted Init via Sparse Matrix-decomposition in Flow-matching](https://mlsanity.com/paper/2609.15643/) (2026)

## Abstract (arXiv)

This paper identifies significant redundancy in the query-key interactions within self-attention mechanisms of diffusion transformer models, particularly during the early stages of denoising diffusion steps. In response to this observation, we present a novel diffusion transformer framework incorporating an additional set of mediator tokens to engage with queries and keys separately. By modulating the number of mediator tokens during the denoising generation phases, our model initiates the denoising process with a precise, non-ambiguous stage and gradually transitions to a phase enriched with detail. Concurrently, integrating mediator tokens simplifies the attention module's complexity to a linear scale, enhancing the efficiency of global attention processes. Additionally, we propose a time-step dynamic mediator token adjustment mechanism that further decreases the required computational FLOPs for generation, simultaneously facilitating the generation of high-quality images within the constraints of varied inference budgets. Extensive experiments demonstrate that the proposed method can improve the generated image quality while also reducing the inference cost of diffusion transformers. When integrated with the recent work SiT, our method achieves a state-of-the-art FID score of 2.01. The source code is available at https://github.com/LeapLabTHU/Attention-Mediators.