Compositional Text-to-Image Generation Via Region-aware Bimodal Direct Preference Optimization

2026-05-27

AI Summary

This work addresses the limitations of current text-to-image generation models in accurately capturing complex compositional semantics—such as attribute binding, object relationships, and counting—by introducing BiDPO, a novel framework that integrates bimodal direct preference optimization with a region-aware guidance mechanism. The authors further construct BiComp, a high-quality compositional preference dataset, enabling effective fine-tuning without requiring additional annotations or architectural modifications. This approach significantly enhances the fine-grained semantic alignment between generated images and intricate textual prompts, achieving substantial improvements over existing methods across multiple benchmarks and markedly improving the fidelity and accuracy of compositional semantics in synthesized imagery.

Abstract

Despite the rapid progress of text-to-image (T2I) models, generating images that accurately reflect complex compositional prompts (covering attribute bindings, object relationships, counting) still remains challenging. To address this, we propose BiDPO, a framework to enhance T2I model's capability of compositional text-to-image generation. We begin by introducing an carefully designed pipeline to construct a large-scale preference dataset, BiComp, with strictly quality control. Then, we extend Diffusion DPO to jointly optimize image and text preferences, which is shown to greatly effective in improving the models to follow complex text prompt in generation. To further enhance the models for fine-grained alignment, we employ a region-level guidance method to focus on regions relevant to compositional concepts. Experimental results demonstrate that our BiDPO substantially improves compositional fidelity, consistently outperforming prior methods across multiple benchmarks. Our approach highlights the potential of preference-based fine-tuning for complex text-to-image tasks, offering a flexible and scalable alternative to existing techniques.

Problem

• compositional text-to-image generation
• attribute binding
• object relationships
• counting
• text-to-image alignment

Innovation

• Compositional Generation
• Direct Preference Optimization
• Region-aware Guidance
• Text-to-Image Synthesis
• Bimodal Alignment