Xianglong Tang

Papers

Total Citations

H-Index

About

Xianglong Tang is a researcher advancing the frontiers of reinforcement learning and scene understanding. His primary contributions lie in developing more efficient goal-generation mechanisms for multi-goal reinforcement learning, a critical challenge in training agents to solve complex, sparse-reward tasks. In his highly cited 2019 work, "Guided goal generation for hindsight multi-goal reinforcement learning" (20 citations), Tang introduced a method that strategically guides the agent toward achievable subgoals, significantly improving learning efficiency. He extended this line of inquiry in 2020 with "Generating attentive goals for prioritized hindsight reinforcement learning" (14 citations), where he incorporated attention mechanisms to prioritize the most informative goals for replay. Beyond reinforcement learning, Tang has also contributed to computer vision, notably with his work on hierarchical inferential methods for indoor scene classification, which addresses the variability in layout and decoration that challenges service robots. His research bridges the gap between autonomous decision-making and perceptual understanding, with his goal-generation techniques offering a practical path toward more capable, sample-efficient AI agents.