Training-Free Object-Background Compositional T2I via Dynamic Spatial Guidance and Multi-Path Pruning

TL;DR

This paper introduces a training-free method for improving object-background compositionality in text-to-image diffusion models by using dynamic spatial guidance and multi-path pruning, enhancing scene coherence.

Contribution

It presents a novel training-free framework with dynamic spatial guidance and multi-path pruning to better model foreground-background interactions in diffusion models.

Findings

Improved background coherence in generated images.

Enhanced object-background compositional alignment.

Consistent performance gains across multiple diffusion backbones.

Abstract

Existing text-to-image diffusion models, while excelling at subject synthesis, exhibit a persistent foreground bias that treats the background as a passive and under-optimized byproduct. This imbalance compromises global scene coherence and constrains compositional control. To address the limitation, we propose a training-free framework that restructures diffusion sampling to explicitly account for foreground-background interactions. Our approach consists of two key components. First, Dynamic Spatial Guidance introduces a soft, time step dependent gating mechanism that modulates foreground and background attention during the diffusion process, enabling spatially balanced generation. Second, Multi-Path Pruning performs multi-path latent exploration and dynamically filters candidate trajectories using both internal attention statistics and external semantic alignment signals, retaining trajectories that better satisfy object-background constraints. We further develop a benchmark specifically designed to evaluate object-background compositionality. Extensive evaluations across multiple diffusion backbones demonstrate consistent improvements in background coherence and object-background compositional alignment.