A Step to Decouple Optimization
in 3DGS

Renjie Ding1, Yaonan Wang1, Min Liu1, Jialin Zhu2, Jiazheng Wang1, Jiahao Zhao2, Wenting Shen2, Feixiang He3, Xiang Chen 1
1National Engineering Research Center of Robot Visual Perception and Control Technology, School of Artificial Intelligence and Robitics, Hunan University
2 Baidu Inc.
3 Central South University
ICLR 2026 Paper arXiv 3DGS-AdamWGS 3DGSMCMC-AdamWGS

Abstract

3D Gaussian Splatting (3DGS) has emerged as a powerful technique for real-time novel view synthesis. As an explicit representation optimized through gradient propagation among primitives, optimization widely accepted in deep neural networks (DNNs) is actually adopted in 3DGS, such as synchronous weight updating and Adam with the adaptive gradient. However, considering the physical significance and specific design in 3DGS, there are two overlooked details in the optimization of 3DGS: (i) update step coupling, which induces optimizer state rescaling and costly attribute updates outside the viewpoints, and (ii) gradient coupling in the moment, which may lead to under- or over-effective regularization. Nevertheless, such a complex coupling is under-explored. After revisiting the optimization of 3DGS, we take a step to decouple it and recompose the process into: Sparse Adam, Re-State Regularization and Decoupled Attribute Regularization. Taking a large number of experiments under the 3DGS and 3DGS-MCMC frameworks, our work provides a deeper understanding of these components. Finally, based on the empirical analysis, we re-design the optimization and propose AdamW-GS by re-coupling the beneficial components, under which better optimization efficiency and representation effectiveness are achieved simultaneously.

BibTeX


        @inproceedings{
        ding2026a,
        title={A Step to Decouple Optimization in 3{DGS}},
        author={Renjie Ding and Yaonan Wang and Min Liu and Jialin Zhu and Jiazheng Wang and Jiahao Zhao and Wenting Shen and Feixiang He and Xiang Chen},
        booktitle={The Fourteenth International Conference on Learning Representations},
        year={2026},
        url={https://openreview.net/forum?id=oapTMDy2Yh}
        }
        @article{ding2026step,
        title={A Step to Decouple Optimization in 3DGS},
        author={Ding, Renjie and Wang, Yaonan and Liu, Min and Zhu, Jialin and Wang, 
          Jiazheng and Zhao, Jiahao and Shen, Wenting and He, Feixiang and Che, Xiang},
        journal={arXiv preprint arXiv:2601.16736},
        year={2026}
      }