IDEA (International Digital Economy Academy)

“GTA-VLA introduces a third input channel — sparse visual cues like a click (affordance point), a drawn box, or a sketched path — that directly conditions the robot's reasoning chain before action is generated.”

“Lei Zhang | IDEA (International Digital Economy Academy) | Corresponding author; head of the research group producing GTA-VLA | leizhang@idea.edu.cn; previously associated with foundational work on T-Rex2 [15] and Rex-Omni [14] for interactive detection”

“Qing Jiang | IDEA | Co-author; previously first author on T-Rex2 and Rex-Omni (interactive detection systems cited in Section 2) | Brings the interactive visual grounding expertise that directly influenced the Guide phase design”

“previously associated with foundational work on T-Rex2 [15] and Rex-Omni [14] for interactive detection — this group has a consistent research thread on interactive visual grounding now extended to robotics”