Everything in Its Place: Benchmarking Spatial Intelligence of Text-to-Image Models
This work addresses a critical limitation in text-to-image generation for applications requiring accurate spatial understanding, though it is incremental as it builds on existing benchmarks and models.
The paper tackles the problem of text-to-image models failing at complex spatial relationships by introducing SpatialGenEval, a benchmark with 1,230 long prompts across 25 scenes, and finds that higher-order spatial reasoning is a primary bottleneck. It also creates the SpatialT2I dataset, which when used for fine-tuning improves performance by up to 5.7% on models like Stable Diffusion-XL, demonstrating a data-centric approach to enhance spatial intelligence.
Text-to-image (T2I) models have achieved remarkable success in generating high-fidelity images, but they often fail in handling complex spatial relationships, e.g., spatial perception, reasoning, or interaction. These critical aspects are largely overlooked by current benchmarks due to their short or information-sparse prompt design. In this paper, we introduce SpatialGenEval, a new benchmark designed to systematically evaluate the spatial intelligence of T2I models, covering two key aspects: (1) SpatialGenEval involves 1,230 long, information-dense prompts across 25 real-world scenes. Each prompt integrates 10 spatial sub-domains and corresponding 10 multi-choice question-answer pairs, ranging from object position and layout to occlusion and causality. Our extensive evaluation of 21 state-of-the-art models reveals that higher-order spatial reasoning remains a primary bottleneck. (2) To demonstrate that the utility of our information-dense design goes beyond simple evaluation, we also construct the SpatialT2I dataset. It contains 15,400 text-image pairs with rewritten prompts to ensure image consistency while preserving information density. Fine-tuned results on current foundation models (i.e., Stable Diffusion-XL, Uniworld-V1, OmniGen2) yield consistent performance gains (+4.2%, +5.7%, +4.4%) and more realistic effects in spatial relations, highlighting a data-centric paradigm to achieve spatial intelligence in T2I models.