Research map / Image generation / Text-to-image diffusion
How Bias Binds: Measuring Hidden Associations for Bias Control in Text-to-Image Compositions
Jeng-Lin Li, Ming-Ching Chang, Wei-Chao Chen
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 BA-Score initialization (ablation) | CBC · Winobias compositional prompts | AFS: 0.68 -> 0.75 | Improved |
| use BA-Score instead of semantic-similarity initialization (ablation) | CBC · Winobias compositional prompts | AFS: 0.64 -> 0.75 | Improved |
| set delta_c to 2 rather than 1 (ablation) | CBC · Winobias compositional prompts | AFS: 0.73 -> 0.75 | Improved |
| set delta_c to 2 rather than 5 (ablation) | CBC · Winobias compositional prompts | AFS: 0.71 -> 0.75 | Improved |
| set delta_r to 0.2 rather than 0.5 (ablation) | CBC · Winobias compositional prompts | FD: 0.12 -> 0.04 | Improved |
| set delta_r to 0.2 rather than 0.3 (ablation) | CBC · Winobias compositional prompts | AFS: 0.74 -> 0.75 | No clear effect |
| use CBC instead of FairQueue | SD-1.5 · Winobias compositional prompts | AFS: 0.60 -> 0.75 | Improved |
| use CBC instead of SD-1.5 | SD-1.5 · Winobias compositional prompts | FD: 0.69 -> 0.04 | Improved |
| use SelfDisc debiasing | SD-1.5 · Winobias compositional prompts | FD: 0.69 -> 0.61 | No clear effect |
| use DGDebias debiasing | SD-1.5 · Winobias compositional prompts | AFS: 0.41 -> 0.35 | Hurt |
| use FairQueue debiasing | SD-1.5 · Winobias compositional prompts | VQA: 0.63 -> 0.45 | Hurt |
| add CBC on plain (non-compositional) prompts | SD-1.5 · Winobias plain prompts | AFS: 0.44 -> 0.80 | 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.
- FairQueue — FairQueue: Rethinking Prompt Learning for Fair Text-to-Image Generation (NeurIPS 2024) · better on 4 of 6 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
- Magnet: We Never Know How Text-to-Image Diffusion Models Work, Until We Learn How Vision-Language Models Function (NeurIPS 2024 · similarity 0.88)
- Beyond Bag-of-Words: Diagnosing Compositional Binding Failures in Vision-Language Models (NeurIPS 2026 · similarity 0.87)
- Token Merging for Training-Free Semantic Binding in Text-to-Image Synthesis (NeurIPS 2024 · similarity 0.87)
- Discovering and Mitigating Visual Biases through Keyword Explanation (CVPR 2024 · similarity 0.86)
- OpenBias: Open-set Bias Detection in Text-to-Image Generative Models (CVPR 2024 · similarity 0.86)
- Association of Objects May Engender Stereotypes: Mitigating Association-Engendered Stereotypes in Text-to-Image Generation (NeurIPS 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.
