Research map / Image generation / Text-to-image diffusion
Guidance Matters: Rethinking the Evaluation Pitfall for Text-to-Image Generation
Dian Xie, Shitong Shao, Lichen Bai, zikai zhou, Bojun Cheng, Shuo Yang, WU JUN, Zeke Xie
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 |
|---|---|---|---|
| raise the CFG scale to the effective scale matched to the method (e-CFG) | CFG · GenEval | overall score (%): 55.84% -> 59.57% | Improved |
| apply Transcendent Diffusion Guidance on top of CFG | CFG · GenEval | overall score (%): 55.84% -> 58.29% | Improved |
| replace TDG with CFG at TDG's effective scale | TDG · GenEval | overall score (%): 58.29% -> 57.23% | No clear effect |
| replace FreeU with CFG at FreeU's effective scale | FreeU · GenEval | overall score (%): 54.22% -> 58.47% | Improved |
| replace APG with CFG at APG's effective scale | APG · GenEval | overall score (%): 55.95% -> 59.57% | Improved |
| replace Z-Sampling with CFG at its effective scale | Z-Sampling · HPD | average winning rate vs CFG (%): 70% -> 62% | Hurt |
| replace CFG++ with CFG at its effective scale | CFG++ · HPD | average winning rate vs CFG (%): 61% -> 53% | Hurt |
| replace TDG with CFG at its effective scale | TDG · HPD | average winning rate vs CFG (%): 56% -> 48% | Hurt |
| replace Z-Sampling with CFG at its effective scale on HPS v2 | Z-Sampling · Pick-a-Pic | HPS v2 winning rate: 90% -> 74% | Hurt |
| replace Z-Sampling with CFG at its effective scale on SD-2.1 | Z-Sampling · Pick-a-Pic (SD-2.1) | HPS v2 winning rate: 90% -> 60% | Hurt |
| increase CFG scale from 5.5 to 20 (ablation) | CFG · Pick-a-Pic (SD-XL) | HPS v2 | Improved |
| increase CFG scale (ablation) | CFG · Pick-a-Pic (SD-XL) | AES | No clear effect |
Closest papers by meaning
- Towards a Golden Classifier-Free Guidance Path via Foresight Fixed Point Iterations (NeurIPS 2025 · similarity 0.91)
- CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models (ICLR 2025 · similarity 0.91)
- Rethinking the Spatial Inconsistency in Classifier-Free Diffusion Guidance (CVPR 2024 · similarity 0.90)
- Rethinking Training for De-biasing Text-to-Image Generation: Unlocking the Potential of Stable Diffusion (CVPR 2025 · similarity 0.90)
- REG: Rectified Gradient Guidance for Conditional Diffusion Models (ICML 2025 · similarity 0.90)
- Elucidating the design space of classifier-guided diffusion generation (ICLR 2024 · similarity 0.90)
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.
