Hu and colleagues investigated how human eye gaze coordinates with full-body movements across a variety of real-world, virtual reality (VR), and augmented reality (AR) scenarios. Prior studies primarily examined eye-head coordination, but this work expanded the analysis to include full-body motion using four public datasets (MoGaze, ADT, GIMO, and EgoBody). Their findings showed that during human-object interactions (pick-and-place), eye gaze strongly correlates with body motion and precedes physical actions, while in human-human interactions (chat, teach), gaze is more closely aligned with the body orientation toward a partner. Building on these insights, the team developed Pose2Gaze, a novel model combining convolutional and spatio-temporal graph convolutional neural networks to predict gaze from head directions and body poses. Compared to state-of-the-art gaze predictors based only on head movements, Pose2Gaze achieved significant accuracy improvements: 24.0% on MoGaze, 10.1% on ADT, 21.3% on GIMO, and 28.6% on EgoBody. Moreover, Pose2Gaze boosted downstream tasks such as eye-based activity recognition, reaching accuracies close to ground truth gaze. These results highlight the rich information available in eye-body coordination and open new avenues for VR/AR gaze prediction research.