Research map / Image generation / Text-to-image diffusion
DOS: Directional Object Separation in Text Embeddings for Multi-Object Image Generation
Dongnam Byun, Jungwon Park, Jungmin Ko, Changin Choi, Wonjong Rhee
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 DOS vectors to CLIP text embeddings | SDXL · Similar Shapes | success rate: 48.00% -> 64.00% | Improved |
| add DOS vectors to CLIP text embeddings | SDXL · Many Objects | success rate: 23.00% -> 48.00% | Improved |
| add DOS vectors to CLIP text embeddings | SDXL · Dissimilar Background Biases | mixture rate: 22.50% -> 17.00% | Improved |
| add DOS vectors to CLIP text embeddings | SD3.5 · Many Objects | success rate: 70.00% -> 76.50% | Improved |
| use DOS instead of TEBOpt | SDXL · four multi-object benchmarks | success rate: 10.50%-24.00% higher SR | Improved |
| use DOS instead of latent modification (A&E, CONFORM) | SDXL · Similar Shapes | inference time: 58.83 s -> 13.87 s (A&E) | Improved |
| apply DOS to object-noun token embeddings only (ablation) | DOS SDXL · Similar Shapes | success rate: 48.0% -> 58.0% | Improved |
| apply DOS to EOT/pooled embeddings only (ablation) | DOS SDXL · Dissimilar Background Biases | mixture rate: 22.5% -> 17.5% | Improved |
| apply DOS to all three embedding types vs one (ablation) | DOS SDXL · Many Objects | success rate: 36.5% -> 48.0% | Improved |
| use adaptive strengths instead of fixed 0.5 (ablation) | DOS SDXL · Many Objects | success rate: 41.0% -> 48.0% | Improved |
| use fixed strength 0.5 for separation vectors (ablation) | DOS SDXL · Similar Shapes | success rate: 48.0% -> 60.5% | Improved |
| add DOS vectors | SDXL · Human preference, Similar Shapes | vote share: 50.52% vs 13.62% (A&E) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CONFORM — CONFORM: Contrast is All You Need for High-Fidelity Text-to-Image Diffusion Models (CVPR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LLM Blueprint: Enabling Text-to-Image Generation with Complex and Detailed Prompts (ICLR 2024 · similarity 0.89)
- Continuous, Subject-Specific Attribute Control in T2I Models by Identifying Semantic Directions (CVPR 2025 · similarity 0.88)
- CLIP Under the Microscope: A Fine-Grained Analysis of Multi-Object Representation (CVPR 2025 · similarity 0.87)
- A Cat Is A Cat (Not A Dog!): Unraveling Information Mix-ups in Text-to-Image Encoders through Causal Analysis and Embedding Optimization (NeurIPS 2024 · similarity 0.87)
- Magnet: We Never Know How Text-to-Image Diffusion Models Work, Until We Learn How Vision-Language Models Function (NeurIPS 2024 · similarity 0.87)
- CONFORM: Contrast is All You Need for High-Fidelity Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.86)
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.
