VIBE: Can a VLM Read the Room?
This addresses the challenge of understanding human social behavior for AI systems, but it is incremental as it focuses on a specific gap in existing VLM capabilities.
The paper tackles the problem of VLMs' limited ability to infer social cues from visual scenes, identifying a Visual Social-Pragmatic Inference gap, and benchmarks several VLMs on a new dataset for this task.
Understanding human social behavior such as recognizing emotions and the social dynamics causing them is an important and challenging problem. While LLMs have made remarkable advances, they are limited to the textual domain and cannot account for the major role that non-verbal cues play in understanding social situations. Vision Language Models (VLMs) can potentially account for this gap, however their ability to make correct inferences over such social cues has received little attention. In this paper, we explore the capabilities of VLMs at social reasoning. We identify a previously overlooked limitation in VLMs: the Visual Social-Pragmatic Inference gap. To target this gap, we propose a new task for VLMs: Visual Social-Pragmatic Inference. We construct a high quality dataset to test the abilities of a VLM for this task and benchmark the performance of several VLMs on it.