Order Is Not Layout: Order-to-Space Bias in Image Generation

Researchers have identified Order-to-Space Bias (OTS) in modern image generation models, where the order entities are mentioned in text prompts incorrectly determines spatial layout and role assignments. The study introduces OTS-Bench to measure this bias and demonstrates that targeted fine-tuning and early-stage interventions can reduce the problem while maintaining generation quality.

• →Order-to-Space Bias causes image generation models to incorrectly map text entity order to spatial positioning, overriding proper visual cues.
• →The bias affects both text-to-image and image-to-image generation models across the industry.
• →Researchers developed OTS-Bench, a new benchmark to systematically measure and evaluate this ordering bias in AI models.
• →The bias is primarily data-driven and occurs during early stages of the image layout formation process.
• →Targeted fine-tuning and early-stage intervention strategies can significantly reduce the bias without compromising image quality.