Evaluating the Generation of Spatial Relations in Text and Image Generative Models
This work addresses the need for comprehensive benchmarking of AI models in spatial reasoning, which is crucial for applications in robotics and human-AI interaction, though it is incremental as it extends existing evaluation methods.
The paper tackled the problem of evaluating spatial relation understanding in generative models by comparing text-to-image (T2I) models and large language models (LLMs) on 10 common prepositions, finding that LLMs are significantly more accurate than T2I models despite being text-trained.
Understanding spatial relations is a crucial cognitive ability for both humans and AI. While current research has predominantly focused on the benchmarking of text-to-image (T2I) models, we propose a more comprehensive evaluation that includes \textit{both} T2I and Large Language Models (LLMs). As spatial relations are naturally understood in a visuo-spatial manner, we develop an approach to convert LLM outputs into an image, thereby allowing us to evaluate both T2I models and LLMs \textit{visually}. We examined the spatial relation understanding of 8 prominent generative models (3 T2I models and 5 LLMs) on a set of 10 common prepositions, as well as assess the feasibility of automatic evaluation methods. Surprisingly, we found that T2I models only achieve subpar performance despite their impressive general image-generation abilities. Even more surprisingly, our results show that LLMs are significantly more accurate than T2I models in generating spatial relations, despite being primarily trained on textual data. We examined reasons for model failures and highlight gaps that can be filled to enable more spatially faithful generations.