In the rapidly evolving landscape of Internet of Things (IoT)-based systems, which are becoming increasingly complex and interconnected, conflict resolution emerges as a pivotal challenge. The proliferation of cohabiting systems that rely on the mutual use of IoT devices results in a corresponding increase in potential conflicts. This requires the development of a highly effective, run-time, and dynamic conflict management framework. Traditional approaches to conflict resolution often struggle to adapt to dynamic and complex environments. This paper introduces a novel attention-based conflict resolution model that dynamically resolves conflicts by focusing on critical system aspects. By integrating real-time contextual and historical data, the model optimizes conflict resolution, enhancing responsiveness and adaptability. Extensive experiments validate the approach, showing significant improvements in addressing IoT conflicts.