AI Paper Atlas

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

MUSE: Multi-Subject Unified Synthesis via Explicit Layout Semantic Expansion

Fei Peng, Junqiang Wu, Yan Li, Tingting Gao, Di Zhang, Huiyuan Fu

ICCV 2025 · Image generation › Text-to-image diffusion · approach: Layout-to-image control · 2 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 decoupled cross-attention with concatenated cross-attention for layout control (ablation)MUSE layout stage on SDXL · MIG Benchlayout success rate (avg): 0.781 -> 0.884Improved
use concatenated cross-attention instead of DCA (ablation)MUSE · MIG BenchCLIP-T: 0.322 -> 0.331Improved
use progressive two-stage training instead of DCA for both layout and subject (ablation)MUSE · MS-Bench-RandomSR-0.65: -0.410 relative to full methodImproved
use progressive two-stage training instead of single-stage CCA+DCA (ablation)MUSE · MS-Bench-RandomCLIP-I-local: -0.085 relative to full methodImproved
train layout first then subject (progressive) instead of jointly in stage 2 (ablation)MUSE · MS-Bench-RandomSR-0.65: -0.043 relative to full methodImproved
train layout control before subject synthesis instead of reversed order (ablation)MUSE · MS-Bench-RandomCLIP-I-local: -0.056 relative to full methodImproved
add bounding box to image tokens rather than concatenating (ablation)MUSE grounding tokens · MS-Bench-RandomCLIP-I-local: -0.015 relative to full methodImproved
increase subject synthesis strength lambda from 0.8 to 1.0 (ablation)MUSE · MS-Bench-RandomCLIP-T: 0.321 -> 0.310Mixed
lower subject synthesis strength lambda from 0.8 to 0.6 (ablation)MUSE · MS-Bench-RandomSR-0.65: 0.755 -> 0.712Hurt
randomize subject layout instead of fixed layoutMS-Diffusion · MS-BenchSR-0.65: 0.706 -> 0.289Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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