In recent years, there has been a proliferation of VR/AR and wearable devices that enable data capture and interaction with the world through an egocentric interface. This has helped develop numerous egocentric robotic interfaces aimed at democratizing robot learning through easy data capture, as demonstrated by projects like ARCap https://stanford-tml.github.io/ARCap, Open & Mobile-TeleVision (https://robot-tv.github.io, https://mobile-tv.github.io), EgoMimic https://egomimic.github.io, etc. Moreover, large-scale egocentric datasets (such as Ego4D, EPIC-KITCHENS, HOT3D, etc.) offer unique advantages for robotics. They facilitate a high-level understanding of tasks, sub-goals, and action predictions based on observed human interaction data. Moreover, since the first-person view aligns naturally with robot-centric perspectives, these datasets can enable the extraction of demonstrations at scale through "passive" human-object interaction data. Another exciting frontier is the development of embodied egocentric world models using such data that can be leveraged for general-purpose robotics. This workshop aims to bring together researchers from robotics, machine learning, and computer vision to discuss the latest advances in developing robotic methods that can effectively leverage egocentric interfaces and egocentric human data for robot learning. We will discuss key challenges, such as how to effectively transfer human interaction data to robotic problems, addressing issues like unstable camera frame movements, lack of action labels, and embodiment gaps between the human and the robot. We will also discuss what interfaces are most effective for demonstrating robot behavior through VR/AR devices. Some key workshop topics include but are not limited to: