Prompt Reinjection: Alleviating Prompt Forgetting in Multimodal Diffusion Transformers for Text-to-Image Generation
On the map: Diffusion models → Image generation, text-to-image and editing
Abstract (arXiv)
Multimodal Diffusion Transformers (MMDiTs) for text-to-image generation maintain separate text and image branches, with bidirectional information flow between text tokens and visual latents throughout denoising. In this setting, we observe a prompt forgetting phenomenon: the semantics of the prompt representation in the text branch is progressively forgotten as depth increases. We further verify this effect on three representative MMDiTs--SD3, SD3.5, and FLUX.1 by probing linguistic attributes of the representations over the layers in the text branch. Motivated by these findings, we introduce a training-free approach, prompt reinjection, which reinjects prompt representations from early layers into later layers to alleviate this forgetting. Experiments on GenEval, DPG, and T2I-CompBench++ show consistent gains in instruction-following capability, along with improvements on metrics capturing preference, aesthetics, and overall text--image generation quality.
Key facts. Cited by 1 paper 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 41 of its 45 references that are in the map. First posted on arXiv 6 Feb 2026.
Builds on
41 of 45 references.
Closest ancestors · 12
Same direction, newest first
- 2025Exploring Multimodal Diffusion Transformers for Enhanced Prompt-based Image Editing11 Aug
- 2025Rethinking Cross-Modal Interaction in Multimodal Diffusion Transformers9 Jun
- 2025BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset14 May
- 2025FreeFlux: Understanding and Exploiting Layer-Specific Roles in RoPE-Based MMDiT for Versatile Image Editing20 Mar
- 2024Stable Flow: Vital Layers for Training-Free Image Editing21 Nov
- 2024Be Yourself: Bounded Attention for Multi-Subject Text-to-Image Generation25 Mar
- 2024ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment8 Mar
- 2023GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment17 Oct
- 2023Linguistic Binding in Diffusion Models: Enhancing Attribute Correspondence through Attention Map Alignment15 Jun
- 2023Attend-and-Excite: Attention-Based Semantic Guidance for Text-to-Image Diffusion Models31 Jan
- 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding23 May
- 2021GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models20 Dec
Other references · 29
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
- 2020Denoising Diffusion Probabilistic Modelscited by 21,076
- 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2020Denoising Diffusion Implicit Modelscited by 9,781
- 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,255
- 2022Scalable Diffusion Models with Transformerscited by 6,054
- 2022Classifier-Free Diffusion Guidancecited by 5,847
- 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
- 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
- 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,761
- 2025Wan: Open and Advanced Large-Scale Video Generative Modelscited by 2,708
17 more
- 2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,045
- 2025Qwen-Image Technical Reportcited by 1,044
- 2023Training Diffusion Models with Reinforcement Learningcited by 982
- 2023Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generationcited by 977
- 2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 890
- 2024Latte: Latent Diffusion Transformer for Video Generationcited by 546
- 2023DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Modelscited by 450
- 2023ELITE: Encoding Visual Concepts into Textual Embeddings for Customized Text-to-Image Generationcited by 416
- 2024The Platonic Representation Hypothesiscited by 382
- 2024SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformerscited by 340
- 2024PixArt-Sigma: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generationcited by 329
- 2023Training-Free Layout Control with Cross-Attention Guidancecited by 313
- 2025Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformercited by 277
- 2023T2I-CompBench++: An Enhanced and Comprehensive Benchmark for Compositional Text-to-image Generationcited by 260
- 2023Fast Training of Diffusion Models with Masked Transformerscited by 176
- 2025Echo-4o: Harnessing the Power of GPT-4o Synthetic Images for Improved Image Generationcited by 100
- 2024Golden Noise for Diffusion Models: A Learning Frameworkcited by 92
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2023Prompt-Free Diffusion: Taking "Text" out of Text-to-Image Diffusion Models
- 2025Rare Text Semantics Were Always There in Your Diffusion Transformer
- 2025Diff-Prompt: Diffusion-Driven Prompt Generator with Mask Supervision
- 2026Rethinking Memorization Mitigation in Diffusion Models: Reinforcing Text Conditioning
- 2025Reusing Computation in Text-to-Image Diffusion for Efficient Generation of Image Sets
- 2026TexTailor: Inference-Time Textual Guidance Tailoring for Multimodal Diffusion Transformers
- 2024Enhancing Semantic Fidelity in Text-to-Image Synthesis: Attention Regulation in Diffusion Models
- 2023Prompting Hard or Hardly Prompting: Prompt Inversion for Text-to-Image Diffusion Models