Do VLMs Have Bad Eyes? Diagnosing Compositional Failures via Mechanistic Interpretability
This addresses a key limitation in VLMs for AI applications requiring robust visual understanding, though it is an incremental diagnostic study.
The paper investigates why Vision-Language Models (VLMs) struggle with compositional generalization and object binding, finding that superposition in CLIP's vision encoder neurons hinders feature representation and reasoning.
Vision-Language Models (VLMs) have shown remarkable performance in integrating visual and textual information for tasks such as image captioning and visual question answering. However, these models struggle with compositional generalization and object binding, which limit their ability to handle novel combinations of objects and their attributes. Our work explores the root causes of these failures using mechanistic interpretability techniques. We show evidence that individual neurons in the MLP layers of CLIP's vision encoder represent multiple features, and this "superposition" directly hinders its compositional feature representation which consequently affects compositional reasoning and object binding capabilities. We hope this study will serve as an initial step toward uncovering the mechanistic roots of compositional failures in VLMs. The code and supporting results can be found https://github.com/Mystic-Slice/Do-VLMs-Have-Bad-Eyes.