Research map / Vision-language and multimodal / Human and embodied modeling
Omni-Attribute: Open-vocabulary Attribute Encoder for Visual Concept Personalization
Tsai-Shien Chen, Aliaksandr Siarohin, Gordon Guocheng Qian, Kuan-Chieh Jackson Wang, Egor Nemchinov, Moayed Haji-Ali, Riza Alp Guler, Willi Menapace, Ivan Skorokhodov, Anil Kag, Jun-Yan Zhu, Sergey Tulyakov
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 contrastive loss (Lcon) (ablation) | Omni-Attribute · 1K validation pairs | cosine-sim gap: -0.002 -> 0.608 | Improved |
| make MLLM trainable via LoRA instead of frozen (ablation) | Omni-Attribute (no Lcon) · Attribute personalization | attribute fidelity: 0.494 -> 0.651 | Improved |
| make MLLM trainable via LoRA instead of frozen (ablation) | Omni-Attribute (no Lcon) · Attribute personalization | text fidelity: 0.920 -> 0.873 | Hurt |
| full finetuning instead of LoRA (ablation) | Omni-Attribute (no Lcon) · Attribute personalization | average score: 0.774 -> 0.747 | Hurt |
| use 8 self-attention layers in connector instead of 1 linear layer (frozen MLLM) (ablation) | Omni-Attribute (no Lcon) · Attribute personalization | average score: 0.745 -> 0.746 | No clear effect |
| increase contrastive weight and temperature (e, g) (ablation) | Omni-Attribute · 1K validation pairs / attribute fidelity | cosine-sim gap vs attribute fidelity: gap 0.738 vs 0.608, Attr-F 0.513 vs 0.641 | Mixed |
| use Omni-Attribute instead of CLIP, DINOv2, Qwen-VL encoders and editing models | Omni-Attribute · 15-attribute personalization benchmark | conditioning fidelity / naturalness (MLLM and human) | Improved |
Closest papers by meaning
- Compositional Caching for Training-free Open-vocabulary Attribute Detection (CVPR 2025 · similarity 0.87)
- Omni-ID: Holistic Identity Representation Designed for Generative Tasks (CVPR 2025 · similarity 0.85)
- GUIDED: Granular Understanding via Identification, Detection, and Discrimination for Fine-Grained Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.84)
- OW-VAP: Visual Attribute Parsing for Open World Object Detection (ICML 2025 · similarity 0.84)
- Unified Personalized Understanding, Generating and Editing (CVPR 2026 · similarity 0.84)
- Tree of Attributes Prompt Learning for Vision-Language Models (ICLR 2025 · similarity 0.84)
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.
