AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuning

Le Zhuo, Liangbing Zhao, Sayak Paul, Yue Liao, Renrui Zhang, Yi Xin, Peng Gao, Mohamed Elhoseiny, Hongsheng Li

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 28 citations (OpenAlex)

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 changedWhereMeasuredEffect
add noise-level scaling (32 samples)FLUX.1-dev · GenEvalOverall: 0.67 -> 0.85Improved
add prompt-level scaling (ablation)FLUX.1-dev + noise scaling · GenEvalOverall: 0.85 -> 0.87Improved
add reflection-level scaling (ablation)FLUX.1-dev + noise & prompt scaling · GenEvalOverall: 0.87 -> 0.91Improved
add reflection-level scaling (ablation)FLUX.1-dev + noise & prompt scaling · GenEvalPosition: 0.80 -> 0.93Improved
add reflection-level scalingFLUX.1-dev + noise scaling · GenEvalAttribute binding: 0.78 -> 0.72Hurt
use sequential refinement (N1 M16) instead of parallel (N16 M1) (ablation)ReflectionFlow · GenEval (GPT-4o verifier, budget 16)Overall: 0.74 -> 0.78Improved
increase refinement depth with fixed budget (N8M2 to N2M8) (ablation)ReflectionFlow · GenEval (GPT-4o verifier, budget 16)Overall: 0.76 -> 0.78Improved
use SANA verifier instead of GPT-4o verifier (ablation)ReflectionFlow · GenEvalOverall: 0.91Improved
use oracle verifier (upper bound) instead of SANA verifier (ablation)ReflectionFlow · GenEvalOverall: 0.91 -> 0.98Improved
apply reflection to hard promptsReflectionFlow · GenEval hard promptscorrectness: 0.10 -> 0.81Improved
apply reflection to medium promptsReflectionFlow · GenEval medium promptscorrectness: 0.55 -> 0.85Improved
apply reflection to easy promptsReflectionFlow · GenEval easy promptscorrectness: 0.95 -> 0.97No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy