CircuitProbe: Dissecting Spatiotemporal Visual Semantics with Circuit Tracing

Abstract

The processing mechanisms underlying language and image understanding in large vision-language models (LVLMs) have been extensively studied. However, the internal reasoning mechanisms of LVLMs for spatiotemporal understanding remain poorly understood. In this work, we introduce a systematic, circuit-based framework designed to investigate how spatiotemporal visual semantics are represented and processed within these LVLMs. Specifically, our framework comprises three circuits: visual auditing circuit, semantic tracing circuit, and attention flow circuit. Through the lens of these circuits, we discover that visual semantics are highly localized to specific object tokens--removing these tokens can degrade model performance by up to 92.6%. Furthermore, we identify that interpretable concepts of objects and actions emerge and become progressively refined in the middle-to-late layers of LVLMs. In contrary to the current works that solely focus on objects in one image, we reveal that the middle-to-late layers of LVLMs exhibit specialized functional localization for spatiotemporal semantics. Our findings offer significant mechanistic insights into spatiotemporal semantics analysis of LVLMs, laying a foundation for designing more robust and interpretable models.