AI Paper Atlas

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

TexTailor: Inference-Time Textual Guidance Tailoring for Multimodal Diffusion Transformers

Binglei Li, Mengping Yang, Zhiyu Tan, Junping Zhang, Hao Li

ECCV 2026 · Image generation › Text-to-image diffusion · approach: Compositional text-to-image guidance · 6 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 block-selective textual hidden-state enhancementSD3.5 · T2I-CompBenchscore: 56.92% to 63.00%Improved
add block-selective textual enhancementSD3.5 · GenEvaloverall: 66.42% to 71.63%Improved
add TexTailor full sentence tailoringFLUX.1-Dev · T2I-CompBench Colorscore: 0.7322 -> 0.7804Improved
add TexTailor to Qwen ImageQwen Image · T2I-CompBench Shapescore: 0.6358 -> 0.6348No clear effect
weaken text states with lambda=0.7 instead of enhancing (ablation)TexTailor on FLUX · T2I-CompBench Colorscore: 0.7322 -> 0.3161Hurt
enhance selected blocks instead of random 5 blocks (ablation)TexTailor on FLUX · T2I-CompBench Colorscore: 0.7624 -> 0.7804Improved
enhance selected blocks instead of all blocks (ablation)TexTailor on FLUX · T2I-CompBench Colorscore: 0.2360 -> 0.7804Improved
increase enhanced block count from 5 to 9 (ablation)TexTailor on FLUX · T2I-CompBench colorscore: 0.7863 -> 0.7073Hurt
raise enhancing scale from 1.5 to 2.0 (ablation)TexTailor on FLUX · T2I-CompBench Colorscore: 0.7804 -> 0.7415Hurt
use token-level instead of sentence-level enhancement (ablation)TexTailor on Qwen Image · T2I-CompBench Amountscore: 0.7359 -> 0.7616Improved
use sentence-level enhancement on counting (ablation)TexTailor on FLUX · GenEval Countscore: 0.6375 -> 0.6000Hurt
add TexTailor under different samplers and steps (ablation)FLUX.1-Dev · T2I-CompBench overallscore: 47.39 -> 50.46Improved

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.

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