Research map / Image generation / Text-to-image diffusion
Customized Generation Reimagined: Fidelity and Editability Harmonized
Jian Jin, Yang Shen, Zhenyong Fu, Jian 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 |
|---|---|---|---|
| replace manual prompt templates with learnable image-specific contexts (ICO) for fine-tuning (ablation) | Stable Diffusion customization · 16 customization subjects (CLIP fidelity-editability curve) | fidelity-editability trade-off | Improved |
| increase concept-specific context length S (ablation) | ICO prompt scheme · customization subset | concept fidelity vs editability | Mixed |
| increase total image-specific context length M+L (ablation) | ICO prompt scheme · customization subset | overall trade-off | Improved |
| add DCI dual-branch integration at inference on ICO fine-tuned model | ICO · 16 customization subjects | concept fidelity and text alignment | Improved |
| add DCI to ICO | ICO · User study vs DreamBooth | text alignment preference: 58.33 % -> 66.67 % | Improved |
| add DCI to ICO | ICO · User study vs Custom Diffusion | visual alignment preference: 57.29 % -> 68.75 % | Improved |
| increase injection strength lambda (ablation) | DCI · qualitative examples | text alignment vs fidelity | Mixed |
| remove aggregation masks (directly add two branches) (ablation) | DCI · qualitative samples | generation quality | Hurt |
| use entire prompt for each branch instead of separated content prompts (ablation) | DCI · qualitative samples | generation quality | Improved |
| aggregate branches at every cross-attention layer (deep aggregation) (ablation) | DCI · qualitative samples | auxiliary branch injection capacity | Improved |
| use DCI for counterfactual contexts such as swimming on the moon | DCI · rare contexts | prompt faithfulness | No clear effect |
Closest papers by meaning
- Customization Assistant for Text-to-Image Generation (CVPR 2024 · similarity 0.89)
- MC^2: Multi-concept Guidance for Customized Multi-concept Generation (CVPR 2025 · similarity 0.89)
- Navigating Text-To-Image Customization: From LyCORIS Fine-Tuning to Model Evaluation (ICLR 2024 · similarity 0.88)
- Direct Consistency Optimization for Robust Customization of Text-to-Image Diffusion models (NeurIPS 2024 · similarity 0.88)
- FreeCustom: Tuning-Free Customized Image Generation for Multi-Concept Composition (CVPR 2024 · similarity 0.87)
- Concept Weaver: Enabling Multi-Concept Fusion in Text-to-Image Models (CVPR 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.
