Research map / Image generation / Text-to-image diffusion
The Intricate Dance of Prompt Complexity, Quality, Diversity and Consistency in T2I Models
Xiaofeng Zhang, Aaron Courville, Michal Drozdzal, Adriana Romero-Soriano
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 |
|---|---|---|---|
| use CFG scale w>1 when generalizing to general prompts (ablation) | toy diffusion model · Gaussian mixture (general-prompt inference) | forward KL: 1.20 -> 23.78 | Hurt |
| use CFG scale w=3 instead of 1 when generalizing to fine-grained prompts (ablation) | toy diffusion model · Gaussian mixture (fine-grained inference) | forward KL: 1.51 -> 0.93 | Improved |
| increase prompt complexity (longer, more specific prompts) | T2I models (LDMv1.5, LDMv3.5L) · CC12M, ImageNet-1k | diversity (Vendi) | Hurt |
| increase prompt complexity | T2I models · CC12M, ImageNet-1k, DCI | consistency (DSG) | Hurt |
| add advanced guidance methods (CADS, interval, APG) instead of vanilla CFG (ablation) | LDMv1.5 and LDMv3.5L · CC12M, ImageNet-1k | diversity (Vendi) | Improved |
| add advanced guidance methods instead of vanilla CFG (ablation) | LDMv1.5 and LDMv3.5L · CC12M, ImageNet-1k | aesthetic quality and consistency | Hurt |
| add LLM prompt expansion (ablation) | LDMv1.5 and LDMv3.5L · CC12M, ImageNet-1k | aesthetic quality | Improved |
| add LLM prompt expansion (ablation) | LDMv1.5 and LDMv3.5L · CC12M, ImageNet-1k | precision and density | Hurt |
| combine prompt expansion with advanced guidance (APG) (ablation) | LDMv3.5L · CC12M | diversity | Improved |
| use newer LDMv3.5L instead of LDMv1.5 | T2I model · CC12M | FDD (reference-based) | Hurt |
Closest papers by meaning
- Who Evaluates the Evaluations? Objectively Scoring Text-to-Image Prompt Coherence Metrics with T2IScoreScore (TS2) (NeurIPS 2024 · similarity 0.88)
- DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models (NeurIPS 2026 · similarity 0.88)
- Minority-Focused Text-to-Image Generation via Prompt Optimization (CVPR 2025 · similarity 0.88)
- ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization (NeurIPS 2024 · similarity 0.87)
- Not All Prompts Are Made Equal: Prompt-based Pruning of Text-to-Image Diffusion Models (ICLR 2025 · similarity 0.87)
- Arena-T2I Hard: Benchmarking and Improving Faithfulness with Dependency-Aware Checklist Rewards (NeurIPS 2026 · similarity 0.87)
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.
