Large Multimodal Agents for Intelligent Transportation: A Systematic Review

A recent review published on arXiv (2608.08184) investigates the current landscape of large multimodal agents (LMAs) within intelligent transportation systems (ITS). It points out a frequent confusion among multimodality, agency, empirical performance, and readiness for deployment. The review analyzes 91 sources, revealing 42 main study families released between January 2023 and August 3, 2026. It presents a taxonomy that differentiates between model-level, system-level, and hybrid multimodality, categorizing each family by system architecture and authority of action. The analysis evaluates evidence through functional capability levels (C0-C3), validation contexts (E0-E4), three propositions (P1-P3), and eight methodological domains (Q1-Q8). Notably, 23 families assess transportation semantics (P1), 24 focus on multidimensional integration (P3), and 19 cover both areas. However, evidence reconciliation (P2) is still unresolved, as no family shows complete evidence provenance. The paper seeks to create an auditable evidence map to clarify LMAs' actual capabilities and readiness for ITS, highlighting the disparity between theoretical systems and their real-world implementation.