Research map / Image generation / Text-to-image diffusion
MoMA: Multimodal LLM Adapter for Fast Personalized Image Generation
Kunpeng Song, Yizhe Zhu, Bingchen Liu, Qing Yan, Ahmed Elgammal, Xiao 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 |
|---|---|---|---|
| add iterative self-attention masking (strength 1.0) in subject-cross-attention (ablation) | MoMA · DreamBooth dataset | CLIP-T (recontextualization): 0.298 -> 0.348 | Improved |
| add iterative self-attention masking (strength 1.0) in subject-cross-attention (ablation) | MoMA · DreamBooth dataset | DINO: 0.575 -> 0.618 | Improved |
| add self-attention masking for texture editing (ablation) | MoMA · DreamBooth dataset | CLIP-T (texture editing): 0.322 (w/o) -> 0.335 (full), 0.231 at half strength | Mixed |
| add subject-cross-attention self-attention feature transfer (beta=1.0) (ablation) | MoMA · DreamBooth dataset | DINO: 0.583 -> 0.618 | Improved |
| add subject-cross-attention feature transfer (ablation) | MoMA · DreamBooth dataset | CLIP-I: 0.782 -> 0.803 | Improved |
| increase subject-cross-attention strength beta from 0.5 to 1.0 (ablation) | MoMA · DreamBooth dataset | CLIP-T (recontextualization): 0.348 -> 0.348 | No clear effect |
| increase subject-cross-attention strength beta from 0.5 to 1.0 for texture editing (ablation) | MoMA · DreamBooth dataset | CLIP-T (texture editing): 0.338 -> 0.335 | Hurt |
| use MoMA multimodal LLM decoder with contextualized features instead of IP-Adapter | IP-Adapter · DreamBooth dataset | CLIP-T (texture editing): 0.319 -> 0.335 | Improved |
| use MoMA instead of IP-Adapter in human evaluation | IP-Adapter · Human evaluation | Prompt Fidelity: 0.351 -> 0.633 | Improved |
Closest papers by meaning
- MM-R1: Unleashing the Power of Unified Multimodal Large Language Models for Personalized Image Generation (AAAI 2026 · similarity 0.89)
- Yo'Chameleon: Personalized Vision and Language Generation (CVPR 2025 · similarity 0.88)
- MANZANO: A Simple and Scalable Unified Multimodal Model with a Hybrid Vision Tokenizer (ICLR 2026 · similarity 0.87)
- Exploring the Deep Fusion of Large Language Models and Diffusion Transformers for Text-to-Image Synthesis (CVPR 2025 · similarity 0.87)
- LMFusion: Adapting Pretrained Language Models for Multimodal Generation (NeurIPS 2025 · similarity 0.87)
- Kosmos-G: Generating Images in Context with Multimodal Large Language Models (ICLR 2024 · similarity 0.87)
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.
