Research map / Image generation / Text-to-image diffusion
X2I: Seamless Integration of Multimodal Understanding into Diffusion Transformer via Attention Distillation
Jian Ma, Qirong Peng, Xu Guo, Chen Chen, Haonan Lu, Zhenyu Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use first and last two layers with average pooling instead of last layer only (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 95.40% -> 96.62% | Improved |
| add adaptive per-layer weights to 3-layer features (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 96.62% -> 97.36% | Improved |
| add custom T5 module to AlignNet (3 layers) (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 97.36% -> 97.57% | No clear effect |
| use all 29 layers instead of 3 with adaptive weights (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 97.36% -> 98.40% | Improved |
| add T5 module to AlignNet on 29 layers (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 98.40% -> 96.73% | Hurt |
| use CNN to aggregate layer features instead of scalar adaptive weights (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 98.40% -> 99.11% | Improved |
| distill attention output instead of block output (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 45.65% -> 99.70% | Improved |
| distill attention output instead of FeedForward output (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 69.58% -> 99.70% | Improved |
| distill attention output instead of second LayerNorm output (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 98.16% -> 99.70% | Improved |
| distill both image and condition sides instead of conditional side only (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 98.13% -> 99.70% | Improved |
| use reverse KL instead of KL (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 99.26% -> 99.70% | Improved |
| use reverse KL instead of JS (ablation) | X2I · Multilingual-General (Qwen2.5-VL) | Performance Ratio vs Flux.1: 99.26% -> 99.70% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- OminiControl — OminiControl: Minimal and Universal Control for Diffusion Transformer (ICCV 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Dual Diffusion for Unified Image Generation and Understanding (CVPR 2025 · similarity 0.90)
- AltDiffusion: A Multilingual Text-to-Image Diffusion Model (AAAI 2024 · similarity 0.89)
- Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for Text-to-Image Synthesis (CVPR 2025 · similarity 0.89)
- De-Diffusion Makes Text a Strong Cross-Modal Interface (CVPR 2024 · similarity 0.88)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (ICML 2025 · similarity 0.88)
- MMaDA: Multimodal Large Diffusion Language Models (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
