Diffusion Transformer

Also known as DiT

Overview

Diffusion Transformers (DiTs) use transformer networks to turn noise into data, typically by repeatedly predicting denoising updates or following a learned flow. Work explores latent- and pixel-space models, conditioning through text, images, geometry, or other modalities, and representation-learning methods that improve generation and downstream perception. To lower the cost of training and inference, researchers use sparse attention, token reduction, caching, quantization, distillation, and hardware-aware designs. A central challenge is balancing speed, memory, and model size against fidelity, temporal or spatial consistency, and reliable performance across tasks and benchmarks.

Latest papers

519 papers46 papers in the last four weeks, down 47% on the four weeks before. 0.3% of all new papers.

Nov 24, 2025

Phase-Aligned RoPE for Mixed-Resolution Diffusion Transformer

Rotary positional embeddings (RoPE) are widely used in diffusion transformers (DiTs) to encode spatial relationships, yet their behavior with mixed-resolution tokens remains underexplored. A natural approach is to rescale token positions from different resolutions into a unified coordinate system before attention, but we show this fails. Our analysis shows that with RoPE, the attention similarity score is a highly structured and periodic function of token distance, so rescaling distances across resolutions moves token pairs to different regions of this periodic function, leading to incorrect attention scores. Motivated by this, we introduce Phase-Aligned Mixed-Resolution Attention (PMA), a training-free mechanism that stabilizes mixed-resolution attention. PMA modifies the RoPE position mapping to enforce a consistent positional scale for every query-key pair, ensuring that relative distances are evaluated under a single reference scale. To further improve local coherence near resolution transitions, we incorporate a lightweight boundary refinement module that softly exchanges features across adjacent scales. Experiments on image and video diffusion models validate our analysis and demonstrate consistent improvements in visual fidelity and computational efficiency.
Nov 14, 2025

DiTGaze: Geometry-Conditioned Diffusion Augmentation for Few-Shot Gaze Estimation

Accurate person-specific gaze estimation benefits from labeled user data, but collecting images that densely cover gaze directions is burdensome. We present DiTGaze, a calibration-efficient augmentation system that follows a two-stream 3D Gaussian Splatting backbone and reformulates its convolutional rendering stage as a geometry-conditioned latent Diffusion Transformer (DiT). A learnable projection and frozen VAE bridge rasterized Gaussian features to latent tokens, while timestep and identity/expression conditioning are integrated through Adaptive Layer Normalization. Intermediate yaw-pitch sampling densifies gaze supervision, and an orthogonality constraint reduces interference between gaze, pose, and expression. On ETH-XGaze, DiTGaze reduces generator-level gaze error by 4.1 percent relative to GazeGaussian. In our runs of the expanded downstream protocol, it ranks first at every calibration budget from one to nine real images; the repeated primary comparison at N = 9 reduces subject-level estimator error by 0.73 degrees and improves 14 of 15 identities (p < 0.001). These results connect the proposed generator changes to statistically supported few-shot personalization gains.
Sep 27, 2025

Stochastic Interpolants via Conditional Dependent Coupling

Existing image generation models face critical challenges regarding the trade-off between computation and fidelity. Specifically, models relying on a pretrained Variational Autoencoder (VAE) suffer from information loss, limited detail, and the inability to support end-to-end training. In contrast, models operating directly in the pixel space incur prohibitive computational cost. Although cascade models can mitigate computational cost, stage-wise separation prevents effective end-to-end optimization, hampers knowledge sharing, and often results in inaccurate distribution learning within each stage. To address these challenges, we introduce a unified multi-stage generative framework formulated under the stochastic interpolant formalism with our Conditional Dependent Coupling strategy as a Flow-Matching variant. It decomposes the generative process into interpolant trajectories at multiple stages, ensuring accurate distribution learning while enabling end-to-end optimization. Importantly, the entire process is modeled as a single unified Diffusion Transformer, eliminating the need for disjoint modules and also enabling knowledge sharing. Under explicitly stated assumptions, we provably reduce both transport cost and asymptotic inference time; we empirically validate the underlying NFE-transport-cost and per-evaluation-cost relations. Extensive experiments demonstrate that our method achieves competitive fidelity and substantially lower wall-clock cost in the low-NFE regime across multiple resolutions.
Sep 26, 2025

RefAM: Attention Magnets for Zero-Shot Referral Segmentation

Most existing approaches to referring segmentation achieve strong performance only through fine-tuning or by composing multiple pre-trained models, often at the cost of additional training and architectural modifications. Meanwhile, large-scale generative diffusion models encode rich semantic information, making them attractive as general-purpose feature extractors. In this work, we introduce a new method that directly exploits features (attention scores) from diffusion transformers for downstream tasks, requiring neither architectural modifications nor additional training. To systematically evaluate these features, we extend benchmarks with vision-language grounding tasks spanning both images and videos. Our key insight is that stop words act as attention magnets: they accumulate surplus attention and can be filtered to reduce noise. Moreover, we identify global attention sinks (GAS) emerging in deeper layers and show that they can be safely suppressed or redirected onto auxiliary tokens, leading to sharper and more accurate grounding maps. We further propose an attention redistribution strategy, where appended stop words partition background activations into smaller clusters, yielding sharper and more localized heatmaps. Building on these findings, we develop RefAM, a simple training-free grounding framework that combines cross-attention maps, GAS handling, and redistribution. Across zero-shot referring image and video segmentation benchmarks, our approach achieves strong performance and surpasses prior methods on most datasets, establishing a new state of the art among training-free methods, without task-specific training or architectural modification. For video, we localize the referent from the spatio-temporal features of a video diffusion transformer via a first-frame query that is then propagated; we therefore do not explicitly resolve temporal (which-frame) expressions.
Aug 20, 2025

Vivid-VR: Distilling Concepts from Text-to-Video Diffusion Transformer for Photorealistic Video Restoration

We present Vivid-VR, a DiT-based generative video restoration method built upon an advanced T2V foundation model, where ControlNet is leveraged to control the generation process, ensuring content consistency. However, conventional fine-tuning of such controllable pipelines frequently suffers from distribution drift due to limitations in imperfect multimodal alignment, resulting in compromised texture realism and temporal coherence. To tackle this challenge, we propose a concept distillation training strategy that utilizes the pretrained T2V model to synthesize training samples with embedded textual concepts, thereby distilling its conceptual understanding to preserve texture and temporal quality. To enhance generation controllability, we redesign the control architecture with two key components: 1) a control feature projector that filters degradation artifacts from input video latents to minimize their propagation through the generation pipeline, and 2) a new ControlNet connector employing a dual-branch design. This connector synergistically combines MLP-based feature mapping with cross-attention mechanism for dynamic control feature retrieval, enabling both content preservation and adaptive control signal modulation. Extensive experiments show that Vivid-VR performs favorably against existing approaches on both synthetic and real-world benchmarks, as well as AIGC videos, achieving impressive texture realism, visual vividness, and temporal consistency. The codes and checkpoints are publicly available at https://github.com/csbhr/Vivid-VR.
Jul 17, 2025

NABLA: Neighborhood Adaptive Block-Level Attention

Recent progress in transformer-based architectures has demonstrated remarkable success in video generation tasks. However, the quadratic complexity of full attention mechanisms remains a critical bottleneck, particularly for high-resolution and long-duration video sequences. In this paper, we propose NABLA, a novel Neighborhood Adaptive Block-Level Attention mechanism that dynamically adapts to sparsity patterns in video diffusion transformers (DiTs). By leveraging block-wise attention with adaptive sparsity-driven threshold, NABLA reduces computational overhead while preserving generative quality. Our method does not require custom low-level operator design and can be seamlessly integrated with PyTorch's Flex Attention operator. Experiments demonstrate that NABLA achieves up to 2.7x faster training and inference compared to baseline almost without compromising quantitative metrics (CLIP score, VBench score, human evaluation score) and visual quality drop. The code and model weights are available here: https://github.com/gen-ai-team/Wan2.1-NABLA
Jun 4, 2025

Generalizable dose denoising using diffusion transformer framework for pencil beam scanning proton therapy

Background and purpose: Intensity-modulated proton therapy (IMPT) enables conformal dose delivery for head and neck (H&N) cancers. However, its sensitivity to inter-fractional anatomical changes motivates online adaptations, where rapid and accurate Monte Carlo (MC) dose calculation remains a major bottleneck. Low-statistics MC simulations reduce computation time but introduce noise. This study aimed to develop a diffusion-transformer framework to recover high-statistics dose distributions from noisy MC simulations. Materials and methods: A diffusion-transformer-based dose denoising framework was developed. Eighty H&N patients were used for training and additional 10 H&N patients and 30 patients with lung, breast, and prostate cancers (10 per site) were used for evaluation. Noisy and high-statistics dose maps were generated using MCsquare. The model was trained using noisy dose maps and CT images as inputs and high-statistics dose maps as ground truth, with mean squared error, residual, regional mean absolute error (MAE), and dose-volume histogram (DVH) losses. Performance was evaluated using MAE, 3D gamma passing rates, and DVH indices. Results: Mean whole-body MAE ranged from 0.08 +/- 0.06 to 0.20 +/- 0.11 Gy (Relative Biological Effectiveness, RBE) (RBE = 1.1) across test cohorts. Mean whole-body 3 %/2 mm gamma passing rates ranged from 95.8 +/- 0.8 % to 97.6 +/- 0.9 %. DVH indices from high-statistics and predicted dose distributions showed close agreement for clinical target volumes and organs of interest. Conclusions: The diffusion-transformer-based framework denoised low-statistics MC dose distributions. Despite being trained on H&N data, the model generalized to lung, breast, and prostate cohorts without retraining.
May 16, 2025

Conditioning Residuals for Diffusion Models via Representation Feedback

Diffusion models now serve as a common foundation for multimedia generation, and useful intermediate representations emerge during their generative training. Standard architectures, however, propagate these representations through the main feature stream, without explicitly reintroducing their encoded semantics to later denoising layers. Meanwhile, such backbones already provide a conditioning pathway for global modulation by predefined inputs. This work examines whether this native pathway can also route internally inferred semantics as evolving, sample-dependent cues. We propose Conditioning Residuals, a lightweight feedback mechanism that converts aggregated features into residuals added to condition embeddings. By feeding back compact feature summaries, it provides adaptive generative guidance and encourages a tighter semantic bottleneck, without external encoders, auxiliary objectives, or sampling-time changes. It supports feedback at one or multiple depths in UNet and DiT backbones, with negligible overhead. Across diffusion formulations, backbone configurations, and datasets, experiments show consistent gains in generative performance, along with stronger representations in downstream linear probing and segmentation. Mechanistic analyses reveal improved generative training dynamics and reshaped feature structure, suggesting a grounded, generalizable way to enhance diffusion backbones from within.
Oct 17, 2024

FiTv2: Scalable and Improved Flexible Vision Transformer for Diffusion Model

Nature is infinitely resolution-free. In the context of this reality, existing diffusion models, such as Diffusion Transformers, often face challenges when processing image resolutions outside of their trained domain. To address this limitation, we conceptualize images as sequences of tokens with dynamic sizes, rather than traditional methods that perceive images as fixed-resolution grids. This perspective enables a flexible training strategy that seamlessly accommodates various aspect ratios during both training and inference, thus promoting resolution generalization and eliminating biases introduced by image cropping. On this basis, we present the Flexible Vision Transformer (FiT), a transformer architecture specifically designed for generating images with unrestricted resolutions and aspect ratios. We further upgrade the FiT to FiTv2 with several innovative designs, includingthe Query-Key vector normalization, the AdaLN-LoRA module, a rectified flow scheduler, and a Logit-Normal sampler. Enhanced by a meticulously adjusted network structure, FiTv2 exhibits 2×2\times convergence speed of FiT. When incorporating advanced training-free extrapolation techniques, FiTv2 demonstrates remarkable adaptability in both resolution extrapolation and diverse resolution generation. Additionally, our exploration of the scalability of the FiTv2 model reveals that larger models exhibit better computational efficiency. Furthermore, we introduce an efficient post-training strategy to adapt a pre-trained model for the high-resolution generation. Comprehensive experiments demonstrate the exceptional performance of FiTv2 across a broad range of resolutions. We have released all the codes and models at https://github.com/whlzy/FiT to promote the exploration of diffusion transformer models for arbitrary-resolution image generation.