ARL HuRDL (Army Research Laboratory Human Robot Dialogue Learning) Corpus Videos is a video collection of 22 human-human interactions in a situated reference resolution task in a virtual environment. In the task, participants played the role of a robot and communicated with a live, remote experimenter through text-based messaging. These data, along with the corresponding dialogue transcripts, enable empirical insights into how humans ask questions to resolve multimodal uncertainty, and can inform algorithms for automated question generation and clarification.