Memory-Augmented Vision-Language Models for Interactive Autonomous Navigation in Dynamic Environments. JAAIR 2026, 5 (1).