Current directions include long-tail object understanding, semantic grounding under ambiguity, and robust multimodal fusion for agents that must act with incomplete observations.
This page is a placeholder for future project details, papers, demos, and datasets.
]]>Current directions include long-tail object understanding, semantic grounding under ambiguity, and robust multimodal fusion for agents that must act with incomplete observations.
]]>We are interested in navigation policies, memory-augmented world models, and planning systems that remain effective when tasks require multi-step reasoning across large spaces.
This page is a placeholder for future project details, papers, demos, and datasets.
]]>YESBUT (v1 & v2) investigates whether multimodal AI systems can comprehend contradictory humor in comics. The work focuses on comparative reasoning: models must connect two related visual situations, identify the apparent agreement implied by “yes,” and then recognize the contradiction introduced by “but.”
Tuo Liang, Zhe Hu, Jing Li, Hao Zhang, Yiren Lu, Yunlai Zhou, Yiran Qiao, Disheng Liu, Jierui Peng, Jing Ma, Yu Yin
Expo-GS studies high dynamic range reconstruction with Gaussian Splatting. By incorporating exposure awareness into a signed distance formulation, the method aims to better handle scenes captured under challenging brightness variation and exposure inconsistency.
Chaoda Song, Yiren Lu, Xinpeng Li, Yunlai Zhou, Yanyan Zhang, Yu Yin, Vipin Chaudhary
Current directions include long-tail object understanding, semantic grounding under ambiguity, and robust multimodal fusion for agents that must act with incomplete observations.
This page is a placeholder for future project details, papers, demos, and datasets.
]]>Praxis-VLM explores vision-grounded decision making through text-driven reinforcement learning. The work studies how language supervision and visual grounding can be combined to improve decision-making behavior in multimodal systems.
Zhe Hu, Jing Li, Zhongzhu Pu, Hou Pong Chan, Yu Yin
Noise Guided Splatting studies the false transparency artifact in 3D Gaussian Splatting. The method injects opaque noise Gaussians into object volumes during training, encouraging surface Gaussians to learn higher opacity and produce cleaner geometry-aware reconstructions.
Aly El Hakie, Yiren Lu, Yu Yin, Michael Jenkins, Yehe Liu
Vision-Language Models (VLMs) have achieved remarkable progress in language-grounded visual understanding, yet a critical gap remains: spatial intelligence. This survey provides a comprehensive overview of recent advances, taxonomies, and evaluations toward building spatially intelligent AI systems.
Disheng Liu, Tuo Liang, Zhe Hu, Jierui Peng, Yiren Lu, Yi Xu, Yun Fu, Yu Yin
Vision-Language Models (VLMs) have achieved remarkable progress in language-grounded visual understanding, yet a critical gap remains: spatial intelligence. This survey provides a comprehensive overview of recent advances, taxonomies, and evaluations toward building spatially intelligent AI systems.
Disheng Liu, Tuo Liang, Zhe Hu, Jierui Peng, Yiren Lu, Yi Xu, Yun Fu, Yu Yin