Research map / Image generation / Text-to-image diffusion
Boost the Identity-Preserving Embedding for Consistent Visual Generation
Zixun Xia, Kai Wang, Shuyu Guo, Boqian Li, jian Yang, Yaxing Wang
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 adaSVR on the [EoT] token in addition to subject tokens (ablation) | BIPE · ConsiStory+ | DreamSim: 0.2631 -> 0.2139 | Improved |
| apply adaSVR only to subject tokens (ablation) | BIPE · ConsiStory+ | CLIP-I: 86.55 | No clear effect |
| add UniK on top of adaSVR (ablation) | BIPE · ConsiStory+ | DreamSim: 0.2139 -> 0.2053 | Improved |
| use UniK alone without adaSVR (ablation) | BIPE · ConsiStory+ | DreamSim: 0.2686 | No clear effect |
| add adaSVR on EoT and subject tokens (full BIPE vs adaSVR on subject only) (ablation) | BIPE · ConsiStory+ | CLIP-I: 86.55 -> 89.10 | Improved |
| full adaSVR + UniK slightly lowers text alignment versus single components (ablation) | BIPE · ConsiStory+ | CLIP-T: 31.79 -> 31.44 | Hurt |
| use BIPE instead of ConsiStory or StoryDiffusion | SDXL · ConsiStory+ | Memory (GB): 46.47 -> 17.16 | Improved |
| switch from Consistent Prompts to Diverse Prompts benchmark | BIPE · DiverStory vs ConsiStory+ | DreamSim: 0.2053 -> 0.2918 | Hurt |
| compare IP-Adapter (trained) to BIPE on identity | IP-Adapter · ConsiStory+ | CLIP-T: 31.44 vs 29.76 | Hurt |
Closest papers by meaning
- CoRe: Context-Regularized Text Embedding Learning for Text-to-Image Personalization (AAAI 2025 · similarity 0.88)
- VideoBooth: Diffusion-based Video Generation with Image Prompts (CVPR 2024 · similarity 0.88)
- One-Prompt-One-Story: Free-Lunch Consistent Text-to-Image Generation Using a Single Prompt (ICLR 2025 · similarity 0.87)
- Harmonizing Visual and Textual Embeddings for Zero-Shot Text-to-Image Customization (AAAI 2025 · similarity 0.87)
- PromptCoT: Align Prompt Distribution via Adapted Chain-of-Thought (CVPR 2024 · similarity 0.87)
- Salient Concept-Aware Generative Data Augmentation (NeurIPS 2025 · 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.
