Is VLA Reasoning Faithful? Probing Safety of Chain-of-Causation in Autonomous Driving Models

Nicanor Mayumu, Xiaoheng Deng, Patrick Mukala

We present the first systematic study of faithfulness in Vision-Language-Action (VLA) driving models, analyzing 300 Alpamayo-R1-10B inferences across 100 diverse PhysicalAI-AV scen…