🏢 Institute of Automation, Chinese Academy of Sciences

OneRef: Unified one-tower model surpasses existing methods in visual grounding and segmentation by leveraging a novel Mask Referring Modeling paradigm.

Researchers enhanced brain recording-based visual reconstruction using a novel Vision Transformer 3D framework integrated with LLMs, achieving superior performance in visual reconstruction, captioning…

RL-instructed language models excel at strategic communication in One Night Ultimate Werewolf, demonstrating the importance of discussion tactics in complex games.

Latent Neural Operator (LNO) dramatically improves solving PDEs by using a latent space, boosting accuracy and reducing computation costs.

Happy: a novel debiased learning framework, excels at continually discovering new categories from unlabeled data while retaining knowledge of previously learned ones, overcoming existing bias issues a…

CP3ER, a novel consistency policy with prioritized proximal experience regularization, significantly boosts sample efficiency and stability in visual reinforcement learning, achieving state-of-the-art…