Adaptive Memory-Augmented Video-Language Models for Fine-Grained Temporal Event Reasoning. JAAIR 2026, 1 (1).