Probing Visual Concepts in Lightweight Vision-Language Models for Automated Driving
This work improves understanding of VLM failure cases for researchers and developers working on automated driving applications, offering insights into perceptual and cognitive limitations.
This paper investigates why Vision-Language Models (VLMs) fail on simple visual questions relevant to automated driving by examining their intermediate activations. They found that concepts like object presence are linearly encoded, while spatial concepts like orientation are only implicitly encoded. They identified two failure modes: perceptual failure, where visual information is not linearly encoded, and cognitive failure, where encoded visual information is not correctly aligned with language semantics.
The use of Vision-Language Models (VLMs) in automated driving applications is becoming increasingly common, with the aim of leveraging their reasoning and generalisation capabilities to handle long tail scenarios. However, these models often fail on simple visual questions that are highly relevant to automated driving, and the reasons behind these failures remain poorly understood. In this work, we examine the intermediate activations of VLMs and assess the extent to which specific visual concepts are linearly encoded, with the goal of identifying bottlenecks in the flow of visual information. Specifically, we create counterfactual image sets that differ only in a targeted visual concept and then train linear probes to distinguish between them using the activations of four state-of-the-art (SOTA) VLMs. Our results show that concepts such as the presence of an object or agent in a scene are explicitly and linearly encoded, whereas other spatial visual concepts, such as the orientation of an object or agent, are only implicitly encoded by the spatial structure retained by the vision encoder. In parallel, we observe that in certain cases, even when a concept is linearly encoded in the model's activations, the model still fails to answer correctly. This leads us to identify two failure modes. The first is perceptual failure, where the visual information required to answer a question is not linearly encoded in the model's activations. The second is cognitive failure, where the visual information is present but the model fails to align it correctly with language semantics. Finally, we show that increasing the distance of the object in question quickly degrades the linear separability of the corresponding visual concept. Overall, our findings improve our understanding of failure cases in VLMs on simple visual tasks that are highly relevant to automated driving.