T2I Diffusion Models

T2I: Text-to-Image

Text-to-image diffusion models turn prompts into images by iteratively denoising a latent or pixel representation, using latent-diffusion U-Nets, diffusion transformers, and rectified-flow models. Researchers improve prompt and spatial control with classifier-free guidance, cross-attention interventions, LoRA adapters, and reference or layout conditioning, while preference learning and reinforcement learning tune outputs to human or task rewards. Another major effort distills models into few-step generators and uses caching, quantization, and latent compression to reduce inference costs; benchmarks test composition, text rendering, image quality, and human preference. A central challenge is balancing prompt fidelity and diversity with speed, quality, and safeguards: realism or alignment gains do not guarantee robust composition, fair representation, or reliable concept removal.

Latest papers

403 papers68 papers in the last four weeks, up 172% on the four weeks before. 0.4% of all new papers.

LESA: Learnable Stage-Aware Predictors for Diffusion Model Acceleration
Feb 24, 2026Peiliang Cai, Jiacheng Liu, Haowen Xu +3Diffusion Model Inference Acceleration Diffusion Models
Shanghai Jiao Tong University
You Don't Need All That Attention: Surgical Memorization Mitigation in Text-to-Image Diffusion Models
Feb 23, 2026Kairan Zhao, Eleni Triantafillou, Peter TriantafillouMemorization in Generative Models T2I Diffusion Models
University of Warwick, UK · Google DeepMind, UK
When Pretty Isn't Useful: Investigating Why Modern Text-to-Image Models Fail as Reliable Training Data Generators
Feb 23, 2026Krzysztof Adamkiewicz, Brian Bernhard Moser, Stanislav Frolov +3Synthetic-to-Real Domain Adaptation Distribution Shift
1RPTU University Kaiserslautern-Landau, Kaiserslautern, Germany · 2German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany
Localized Concept Erasure in Text-to-Image Diffusion Models via High-Level Representation Misdirection
Feb 23, 2026Uichan Lee, Jeonghyeon Kim, Sangheum HwangDiffusion Model Unlearning T2I Diffusion Models
Department of Data Science, Seoul National University of Science and Technology
ConceptPrism: Concept Disentanglement in Personalized Diffusion Models via Residual Token Optimization
Feb 23, 2026Minseo Kim, Minchan Kwon, Dongyeun Lee +2T2I Generation Subject-Driven Image Generation
KAIST, South Korea · Hanbat National University, South Korea
ChordEdit: One-Step Low-Energy Transport for Image Editing
Feb 22, 2026Liangsi Lu, Xuhang Chen, Minzhe Guo +3Text-Guided Image Editing Diffusion-Based Image Editing
Guangdong University of Technology · Huizhou University · Shenzhen University +1
TokenTrace: Multi-Concept Attribution through Watermarked Token Recovery
Feb 22, 2026Li Zhang, Shruti Agarwal, John Collomosse +2Diffusion Model Watermarking Robust Watermarking
Adobe Research · University of California, San Diego · DECaDE, University of Surrey
When Backdoors Go Beyond Triggers: Semantic Drift in Diffusion Models Under Encoder Attacks
Feb 21, 2026Shenyang Chen, Liuwan ZhuAdversarial Attacks Representation Drift
Google · Electrical and Computer Engineering Department, University of Hawai‘i at Mānoa
CRAFT-LoRA: Content-Style Personalization via Rank-Constrained Adaptation and Training-Free Fusion
Feb 21, 2026Yu Li, Yujun Cai, Chi ZhangImage Generation T2I Diffusion Models
AGI Lab, Westlake University · George Washington University · The University of Queensland
Rethinking Preference Alignment for Diffusion Models with Classifier-Free Guidance
Feb 21, 2026Zhou Jiang, Yandong Wen, Zhen LiuDiffusion Model Alignment Preference Alignment
Westlake University · The Chinese University of Hong Kong, Shenzhen
Morphological Addressing of Identity Basins in Text-to-Image Diffusion Models
Feb 20, 2026Andrew FraserT2I Generation T2I Diffusion Models
GASS: Geometry-Aware Spherical Sampling for Disentangled Diversity Enhancement in Text-to-Image Generation
Feb 19, 2026Ye Zhu, Kaleb S. Newman, Johannes F. Lutzeyer +3T2I Generation T2I Diffusion Models
Laboratoire d’Informatique (LIX), CNRS, École Polytechnique, IPP, France · Department of Computer Science, Princeton University, USA · FAIR at Meta - Montreal, Canada +3
Amber-Image: Efficient Compression of Large-Scale Diffusion Transformers
Feb 19, 2026Chaojie Yang, Tian Li, Yue Zhang +1Diffusion Model Compression T2I Diffusion Models
Computational Intelligence Dept, HelloGroup Inc.
Efficient Text-Guided Convolutional Adapter for the Diffusion Model
Feb 16, 2026Aryan Das, Koushik Biswas, Swalpa Kumar Roy +2Diffusion Model Adaptation Conditional Image Generation
VIT Bhopal · IIIT Delhi · Tezpur University Assam +1
ELROND: Exploring and decomposing intrinsic capabilities of diffusion models
Feb 10, 2026Paweł Skierś, Emilia Kaczmarczyk, Tomasz Trzciński +1Diffusion Models T2I Diffusion Models
Warsaw University of Technology · IDEAS Research Institute
From Concept Erasure to Style Purification: Contrastive Eigenbases for Artist Style Protection
Feb 8, 2026Tong Zhang, Ru Zhang, Jianyi LiuConcept Erasure T2I Diffusion Models
Beijing University of Posts and Telecommunications, School of Cyberspace Security, Beijing, 100876
TexTailor: Inference-Time Textual Guidance Tailoring for Multimodal Diffusion Transformers
Jan 5, 2026Binglei Li, Mengping Yang, Zhiyu Tan +2Diffusion Transformer Compositional T2I Generation
Fudan University, Shanghai 200433, China · Shanghai Academy of AI for Science, Shanghai 200030, China · Shanghai Innovation Institute, Shanghai 200030, China
Direct Diffusion Score Preference Optimization via Stepwise Contrastive Policy-Pair Supervision
Dec 29, 2025Dohyun Kim, Seungwoo Lyu, Seung Wook Kim +1Diffusion Model Alignment Diffusion Model Fine-Tuning
Dept. of CSE, Korea University · 2NVIDIA
MixFlow Training: Alleviating Exposure Bias with Slowed Interpolation Mixture
Dec 22, 2025Hui Li, Fu-Yun Wang, Haoyuan Xia +4Diffusion Model Fine-Tuning Diffusion-Based Image Generation
Shanghai Innovation Institute · Fudan University · Shanghai Academy of AI for Science +2
Text-Conditioned Background Generation for Editable Multi-Layer Documents
Dec 19, 2025Taewon Kang, Joseph K J, Chris Tensmeyer +4Diffusion-Based Image Editing T2I Diffusion Models
University of Maryland at College Park, United States · Adobe Research, United States
GRAN-TED: Generating Robust, Aligned, and Nuanced Text Embedding for Diffusion Models
Dec 17, 2025Bozhou Li, Sihan Yang, Yushuo Guan +6Text Embedding Evaluation T2I Generation
Peking University · Kling Team, Kuaishou Technology · Xi’an Jiaotong University +1
Dynamic Image Prompt Adapter for Scalable Zero-shot Personalized Text-to-Image Generation
Dec 10, 2025Zhizhong Wang, Tianyi Chu, Zeyi Huang +2Subject-Driven Image Generation T2I Generation
Central Media Technology Institute, Huawei
Locality-Aware Continual Unlearning for Diffusion Models
Dec 2, 2025Naveen George, Naoki Murata, Yuhta Takida +2Diffusion Model Unlearning Machine Unlearning
Indian Institute of Technology Hyderabad, India · Sony AI, Japan · Sony Group Corporation, Japan
Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer
Nov 27, 2025Z-Image Team, Huanqia Cai, Sihan Cao +21Diffusion Transformer Diffusion Model Inference Acceleration
Z-Image Team, Alibaba Group
Anchor to Expand: Semantic Anchoring for Personalized Text-to-Image Diffusion Models
Nov 27, 2025Seoyun Yang, Gihoon Kim, Taesup KimSubject-Driven Image Generation T2I Generation
Graduate School of Data Science, Seoul National University
SONIC: Spectral Optimization of Noise for Inpainting with Consistency
Nov 25, 2025Seungyeon Baek, Erqun Dong, Shadan Namazifard +2Diffusion Model Inversion Image Inpainting
University of British Columbia, Canada · Google DeepMind, USA
TPSO: Training-Free Diverse Image Generation via Semantic Prompt Embedding Optimization
Nov 25, 2025Debin Meng, Chen Jin, Zheng Gao +3T2I Generation T2I Diffusion Models
Queen Mary University of London · Centre for AI, AstraZeneca, UK · University of Bedfordshire +1
Rethinking Robust Adversarial Concept Erasure in Diffusion Models
Oct 31, 2025Qinghong Yin, Yu Tian, Heming Yang +5Adversarial Robustness Diffusion Model Unlearning
Beijing University of Posts and Telecommunications, China · University of Chinese Academy of Sciences, China · Nanjing University of Aeronautics and Astronautics, China
FreeFuse: Multi-Subject LoRA Fusion via Adaptive Token-Level Routing at Test Time
Oct 27, 2025Yaoli Liu, Yao-Xiang Ding, Kun ZhouSubject-Driven Image Generation Compositional T2I Generation
State Key Laboratory of CAD&CG Zhejiang University · Zhejiang University
UltraImageGen: Efficient Ultra-High-Resolution Image Generation with Hierarchical Local Attention
Oct 18, 2025Yuyao Zhang, Yu-Wing TaiImage Generation Diffusion Model Inference Acceleration
Dartmouth College, USA

Feb 24, 2026J/K move · Enter open · S save

LESA: Learnable Stage-Aware Predictors for Diffusion Model Acceleration

Peiliang Cai, Jiacheng Liu, Haowen Xu, Xinyu Wang+2

Shanghai Jiao Tong University

Diffusion Model Inference Acceleration Diffusion Models Diffusion Model Caching T2I Diffusion Models Diffusion Transformer Video Diffusion Models

Open paper AskPDF arXiv

Diffusion models have achieved remarkable success in image and video generation tasks. However, the high computational demands of Diffusion Transformers (DiTs) pose a significant challenge to their practical deployment. While feature caching is a promising acceleration strategy, existing methods based on simple reusing or training-free forecasting struggle to adapt to the complex, stage-dependent dynamics of the diffusion process, often resulting in quality degradation and failing to maintain consistency with the standard denoising process. To address this, we propose a LEarnable Stage-Aware (LESA) predictor framework based on two-stage training. Our approach leverages a Kolmogorov-Arnold Network (KAN) to accurately learn temporal feature mappings from data. We further introduce a multi-stage, multi-expert architecture that assigns specialized predictors to different noise-level stages, enabling more precise and robust feature forecasting. Extensive experiments show our method achieves significant acceleration while maintaining high-fidelity generation. Experiments demonstrate 5.00x acceleration on FLUX.1-dev with minimal quality degradation (1.0% drop), 6.25x speedup on Qwen-Image with a 20.2% quality improvement over the previous SOTA (TaylorSeer), and 5.00x acceleration on HunyuanVideo with a 24.7% PSNR improvement over TaylorSeer. State-of-the-art performance on both text-to-image and text-to-video synthesis validates the effectiveness and generalization capability of our training-based framework across different models. Our code is available at https://github.com/caipeiliang2004/LESA.