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

TL;DR

Modern image generation models exhibit a pervasive Order-to-Space Bias where the order of entities in text influences spatial layout, often leading to incorrect images, but targeted interventions can mitigate this bias.

Contribution

This paper identifies and quantifies the Order-to-Space Bias in image generation models and proposes methods to reduce it without harming quality.

Findings

OTS is widespread in current models.

OTS primarily arises from training data and early layout formation.

Interventions can significantly reduce OTS while maintaining quality.

Abstract

We study a systematic bias in modern image generation models: the mention order of entities in text spuriously determines spatial layout and entity--role binding. We term this phenomenon Order-to-Space Bias (OTS) and show that it arises in both text-to-image and image-to-image generation, often overriding grounded cues and causing incorrect layouts or swapped assignments. To quantify OTS, we introduce OTS-Bench, which isolates order effects with paired prompts differing only in entity order and evaluates models along two dimensions: homogenization and correctness. Experiments show that Order-to-Space Bias (OTS) is widespread in modern image generation models, and provide evidence that it is primarily data-driven and manifests during the early stages of layout formation. Motivated by this insight, we show that both targeted fine-tuning and early-stage intervention strategies can substantially reduce OTS, while preserving generation quality.