When Discourse Pressures Conflict: Information Structure in Vision-Language Model Outputs
For researchers evaluating VLMs, this work highlights a previously overlooked dimension—discourse-appropriate form—beyond content accuracy.
This paper investigates whether vision-language models (VLMs) produce discourse-appropriate information structure (IS) in visually grounded question answering, using Hungarian as a test case. Comparing six VLMs with humans, they find that models over-regularize IS sensitivity, collapsing onto narrow response templates rather than using variable strategies like humans.
Vision-language models (VLMs) are increasingly evaluated for whether they identify the right visual content, but little is known about whether they express such content in a discourse-appropriate form. We address this research gap using information structure (IS), testing whether VLMs distinguish discourse-old Topics from discourse-new Foci in visually grounded question answering. We exploit Hungarian, a language in which Topic and Focus map onto dedicated syntactic positions, making IS choices observable in text. Comparing six VLMs with human participants, we find that models produce IS-relevant constructions, but over-regularise this sensitivity. Under the interacting pressures of discourse status, grammatical role (preference for subject Topics) and definiteness (preference for indefinite Foci), humans choose variable strategies for IS realisation. VLMs, by contrast, collapse onto narrow response templates, resembling mode collapse (Kirk et al., 2024). Our findings suggest that VLM evaluation should look beyond content accuracy to how content is packaged for the discourse.