AI Paper Atlas

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

Reflect-DiT: Inference-Time Scaling for Text-to-Image Diffusion Transformers via In-Context Reflection

Shufan Li, Konstantinos Kallidromitis, Akash Gokul, Arsh Koneru, Yusuke Kato, Kazuki Kozuka, Aditya Grover

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Inference-time T2I improvement · 30 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
replace best-of-20 sampling with in-context reflection (Reflect-DiT)SANA-1.0-1.6B · GenEval overallaccuracy: 0.62 -> 0.81Improved
replace best-of-20 with Reflect-DiT at same sample budgetSANA-1.0-1.6B · GenEval overallaccuracy: 0.75 -> 0.81Improved
replace best-of-20 with Reflect-DiTSANA-1.0-1.6B · GenEval (Position)accuracy: 0.54 -> 0.66Improved
use Reflect-DiT with 20 samples instead of SANA-1.5-4.8B best-of-2048SANA-1.0-1.6B · GenEval overallaccuracy: 0.80 -> 0.81Improved
replace SFT or DPO fine-tuning plus best-of-N with Reflect-DiTSANA-1.0-1.6B · GenEval overall, N=4accuracy: 0.736 -> 0.776Improved
replace Diffusion-DPO with Reflect-DiTSANA-1.0-1.6B · GenEval overall, N=20accuracy: 0.765 -> 0.807Improved
increase in-context feedback from K=1 to K=3 (ablation)Reflect-DiT · GenEval overall, N=20accuracy: 0.766 -> 0.807Improved
increase in-context feedback from K=3 to K=5 (ablation)Reflect-DiT · GenEval overall, N=20accuracy: 0.807 -> 0.816Improved
increase image embedding size from 4x4 to 8x8 (ablation)Reflect-DiT · GenEval overall, N=20accuracy: 0.786 -> 0.807Improved
increase image embedding size from 6x6 to 8x8 (ablation)Reflect-DiT · GenEval overall, N=20accuracy: 0.801 -> 0.807Improved
replace best-of-20 with Reflect-DiTSANA-1.0 · DPG-Bench Hard-56score: 41.0 -> 51.8Improved
replace best-of-20 with Reflect-DiTSANA-1.0 · DPG-Bench Allscore: 85.6 -> 86.1Improved

Later papers that compared with this method

1 later accepted paper reports Reflect-DiT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

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

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