Research map / Image generation / Text-to-image diffusion
Resolving the Identity Crisis in Text-to-Image Generation
Shubhankar Borse, Farzad Farhadzadeh, Munawar Hayat, Fatih Porikli
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 |
|---|---|---|---|
| fine-tune with DISCO (RL with diversity rewards) | FLUX-Dev · DiverseHumans | Unique Face Accuracy: 48.2 -> 98.6 | Improved |
| fine-tune with DISCO | Krea-Dev · DiverseHumans | Global Identity Spread: 50.6 -> 90.6 | Improved |
| fine-tune with DISCO | FLUX-Dev · MultiHuman-TestBench | Count Accuracy: 61.8 -> 86.6 | Improved |
| fine-tune with DISCO vs GPT-Image-1 | FLUX-Dev · DiverseHumans | Unique Face Accuracy: 85.1 -> 98.6 | Improved |
| fine-tune with DISCO vs GPT-Image-1 | FLUX-Dev · MultiHuman-TestBench | Count Accuracy: 90.7 vs 86.6 | Hurt |
| add group-wise diversity reward to intra-image reward (ablation) | DISCO (Krea) · DiverseHumans | Global Identity Spread: 50.8 -> 72.5 | Improved |
| add group-wise diversity reward to intra-image reward (ablation) | DISCO (Krea) · DiverseHumans | Count Accuracy: 66.2 -> 67.3 | No clear effect |
| add count reward (ablation) | DISCO (Krea) · DiverseHumans | Count Accuracy: 67.3 -> 81.1 | Improved |
| add count reward (ablation) | DISCO (Krea) · DiverseHumans | Global Identity Spread: 72.5 -> 68.3 | Hurt |
| add HPS quality reward (ablation) | DISCO (Krea) · DiverseHumans | HPS score: 31.9 -> 32.4 | Improved |
| add single-stage curriculum (ablation) | DISCO (Krea) · DiverseHumans | Global Identity Spread: 73.7 -> 90.6 | Improved |
| add single-stage curriculum (ablation) | DISCO (Krea) · DiverseHumans | Unique Face Accuracy: 82.6 -> 89.7 | Improved |
Closest papers by meaning
- MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans (NeurIPS 2025 · similarity 0.87)
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (CVPR 2024 · similarity 0.86)
- FaithfulFaces: Pose-Faithful Facial Identity Preservation for Text-to-Video Generation (NeurIPS 2026 · similarity 0.86)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.86)
- CHATS: Combining Human-Aligned Optimization and Test-Time Sampling for Text-to-Image Generation (ICML 2025 · similarity 0.85)
- $\text{ID}^3$: Identity-Preserving-yet-Diversified Diffusion Models for Synthetic Face Recognition (NeurIPS 2024 · 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.
