2026
MangoBench: A Benchmark for Multi-Agent Goal-Conditioned Offline Reinforcement Learning
Yi Wang, Ningze Zhong, Zhiheng Fu, Longguang Wang, Ye Zhang, Yulan Guo
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026
MangoBench, the first benchmark tailored for Goal-Conditioned Offline MARL, covering 3 environments, 4 agent types, and 47 tasks, designed to assess joint-control locomotion, synchronous and asynchronous bimanual manipulation, and robustness to high-dimensional inputs.
[Project Page] [Locomotion Code] [Manipulation Code] [Checkpoints]
MangoBench: A Benchmark for Multi-Agent Goal-Conditioned Offline Reinforcement Learning
Yi Wang, Ningze Zhong, Zhiheng Fu, Longguang Wang, Ye Zhang, Yulan Guo
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026
MangoBench, the first benchmark tailored for Goal-Conditioned Offline MARL, covering 3 environments, 4 agent types, and 47 tasks, designed to assess joint-control locomotion, synchronous and asynchronous bimanual manipulation, and robustness to high-dimensional inputs.
[Project Page] [Locomotion Code] [Manipulation Code] [Checkpoints]
Leveraging Suboptimal and Noisy Trajectories for Goal-Conditional Offline RL
Ningze Zhong, Yi Wang, Bo Wu
ICLR 2026 RSI Workshop 2026
This paper demonstrates that imperfect trajectories in offline goal-conditioned reinforcement learning (OGCRL), typically discarded as harmful, can be leveraged as a valuable source of exploration, enhancing state-space coverage and improving policy learning, especially in complex environments.
Leveraging Suboptimal and Noisy Trajectories for Goal-Conditional Offline RL
Ningze Zhong, Yi Wang, Bo Wu
ICLR 2026 RSI Workshop 2026
This paper demonstrates that imperfect trajectories in offline goal-conditioned reinforcement learning (OGCRL), typically discarded as harmful, can be leveraged as a valuable source of exploration, enhancing state-space coverage and improving policy learning, especially in complex environments.
2024
Tangram-Splatting: Optimizing 3D Gaussian Splatting Through Tangram-inspired Shape Priors
Yi Wang*, Ningze Zhong*, Minglin Chen, Longguang Wang, Yulan Guo (* equal contribution)
ACM Multimedia 2024 (ACM MM) 2024
This study introduces Tangram-Splatting, a novel 3D scene reconstruction method inspired by the tangram puzzle. This method optimizes 3D Gaussian Splatting by diversifying Gaussian functions, achieving a 62.4% reduction in memory overhead while maintaining competitive PSNR performance.
Tangram-Splatting: Optimizing 3D Gaussian Splatting Through Tangram-inspired Shape Priors
Yi Wang*, Ningze Zhong*, Minglin Chen, Longguang Wang, Yulan Guo (* equal contribution)
ACM Multimedia 2024 (ACM MM) 2024
This study introduces Tangram-Splatting, a novel 3D scene reconstruction method inspired by the tangram puzzle. This method optimizes 3D Gaussian Splatting by diversifying Gaussian functions, achieving a 62.4% reduction in memory overhead while maintaining competitive PSNR performance.
2023
CASIT: Collective Intelligent Agent System for Internet of Things
Ningze Zhong*, Yi Wang*, etc (* equal contribution)
IEEE Internet of Things Journal 2023
This article introduces CASIT, a pioneering collective intelligent agent system for IoT, leveraging multiple LLM-based agents with Memory and Summary Mechanisms to collaboratively solve complex tasks and optimize information transmission.
CASIT: Collective Intelligent Agent System for Internet of Things
Ningze Zhong*, Yi Wang*, etc (* equal contribution)
IEEE Internet of Things Journal 2023
This article introduces CASIT, a pioneering collective intelligent agent system for IoT, leveraging multiple LLM-based agents with Memory and Summary Mechanisms to collaboratively solve complex tasks and optimize information transmission.