Research map / Image generation / Text-to-image diffusion
Rethinking Prompt Design for Inference-time Scaling in Text-to-Visual Generation
Subin Kim, Sangwoo Mo, Mamshad Nayeem Rizve, Yiran Xu, Difan Liu, Jinwoo Shin, Tobias Hinz
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 |
|---|---|---|---|
| revise prompt from common failures instead of per-sample failures (ablation) | PRIS · T2V prompt revision ablation | score: 0.619 -> 0.754 | Improved |
| apply per-sample prompt revision instead of no revision (ablation) | PRIS · T2V prompt revision ablation | score: 0.711 -> 0.619 | Hurt |
| revise prompt from common failures instead of per-sample failures (ablation) | PRIS · T2I prompt revision ablation | score: 0.853 -> 0.854 | No clear effect |
| add element-level factual correction instead of decomposed binary VQA (ablation) | EFC verifier · Verifier benchmark (410 prompts) | accuracy: 0.700 -> 0.763 | Improved |
| add PRIS prompt redesign to Best-of-N | FLUX.1-dev · GenAI-Bench | VQAScore: 0.783 -> 0.854 | Improved |
| add PRIS prompt redesign to Best-of-N* on Wan2.1-14B | Wan2.1-14B · VBench-2.0 Controllability & Creativity | average score: 58.51 -> 68.85 | Improved |
| add PRIS to Wan2.1-1.3B vs BoN* on commonsense and physics | Wan2.1-1.3B · VBench-2.0 Commonsense & Physics | average score: 62.61 -> 66.07 | Improved |
| add PRIS on Thermotics dimension | Wan2.1-1.3B · VBench-2.0 Thermotics | score: 86.25 -> 70.37 | Hurt |
| use PRIS with fewer NFEs than BoN under matched wall-clock time | FLUX.1-dev · GenAI-Bench matched compute | score: 0.790 -> 0.834 | Improved |
| use PRIS with fewer NFEs than BoN under matched wall-clock time | Wan2.1-1.3B · T2V matched compute | score: 0.935 -> 0.964 | Improved |
| add standard prompt expansion to FLUX.1-dev | FLUX.1-dev · GenAI-Bench | VQAScore: 0.718 -> 0.769 | Improved |
| add Best-of-N to Wan2.1-1.3B | Wan2.1-1.3B · VBench-2.0 Commonsense & Physics | average score: 63.22 -> 62.61 | Hurt |
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.
- VisionReward — VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation (AAAI 2026) · better on 1 of 1 shared result
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
- PromptCoT: Align Prompt Distribution via Adapted Chain-of-Thought (CVPR 2024 · similarity 0.91)
- Tailored Visions: Enhancing Text-to-Image Generation with Personalized Prompt Rewriting (CVPR 2024 · similarity 0.89)
- Imagine Before You Draw: Visual Prompt Engineering for Image Generation (NeurIPS 2026 · similarity 0.89)
- Visually Guided Decoding: Gradient-Free Hard Prompt Inversion with Language Models (ICLR 2025 · similarity 0.89)
- DynT2I-Eval: A Dynamic Evaluation Framework for Text-to-Image Models (NeurIPS 2026 · similarity 0.88)
- A User-Friendly Framework for Generating Model-Preferred Prompts in Text-to-Image Synthesis (AAAI 2024 · similarity 0.88)
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.
