Research map / Detection and segmentation / Open-set and few-shot detection
UniVerse: A Unified Modulation Framework for Segmentation-Free, Disentangled Multi-Concept Personalization
Quynh Phung, Sandesh Ghimire, Minsi Hu, Chung-Chi Tsai, Jia-Bin Huang
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 RCE pretraining stage (segmentation-supervised FiLM) (ablation) | UniVerse · UniVerseBench (multi-subject) | Avg score: 47.89 -> 48.64 | Improved |
| add RCE pretraining stage (segmentation-supervised FiLM) (ablation) | UniVerse · UniVerseBench (multi-subject) | IP-S: 41.82 -> 42.29 | Improved |
| add RCE pretraining stage (segmentation-supervised FiLM) (ablation) | UniVerse · UniVerseBench (multi-subject) | AES: 53.96 -> 54.98 | Improved |
| add cross-reference augmentation during training (ablation) | UniVerse · UniVerseBench (multi-subject) | Avg score: 47.90 -> 48.64 | Improved |
| add cross-reference augmentation during training (ablation) | UniVerse · UniVerseBench (multi-subject) | IP-S: 41.95 -> 42.29 | Improved |
| add cross-reference augmentation during training (ablation) | UniVerse · UniVerseBench (multi-subject) | AES: 53.84 -> 54.98 | Improved |
| add visual reference latents at inference (ablation) | UniVerse · UniVerseBench (multi-subject) | Avg score: 47.14 -> 48.64 | Improved |
| add visual reference latents at inference (ablation) | UniVerse · UniVerseBench (multi-subject) | IP-S: 40.15 -> 42.29 | Improved |
| add visual reference latents at inference (ablation) | UniVerse · UniVerseBench (multi-subject) | AES: 54.12 -> 54.98 | Mixed |
| UniVerse versus XVerse (second best) on single-subject | XVerse · XVerseBench (single-subject) | Avg score: 74.36 -> 78.14 | Improved |
| UniVerse versus XVerse on multi-subject | XVerse · XVerseBench (multi-subject) | Avg score: 67.93 -> 70.18 | Improved |
| UniVerse versus XVerse on ID-S multi-subject | XVerse · XVerseBench (multi-subject) | ID-S: 67.15 -> 66.69 | No clear effect |
Closest papers by meaning
- ConceptAttention: Diffusion Transformers Learn Highly Interpretable Features (ICML 2025 · similarity 0.88)
- Attention Calibration for Disentangled Text-to-Image Personalization (CVPR 2024 · similarity 0.86)
- LoRAShop: Training-Free Multi-Concept Image Generation and Editing with Rectified Flow Transformers (NeurIPS 2025 · similarity 0.85)
- Factorized Diffusion Autoencoder for Unsupervised Disentangled Representation Learning (AAAI 2024 · similarity 0.85)
- Bridging the gap to real-world language-grounded visual concept learning (NeurIPS 2025 · similarity 0.85)
- Concept Conductor: Orchestrating Multiple Personalized Concepts in Text-to-Image Synthesis (AAAI 2025 · similarity 0.85)
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.
