Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
GauDP: Reinventing Multi-Agent Collaboration through Gaussian-Image Synergy in Diffusion Policies
Authors: Ziye Wang, Li Kang, Yiran Qin, Jiahua Ma, zhanglin peng, LEI BAI, Ruimao Zhang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate Gau DP on the Robo Factory benchmark, which includes diverse multiarm manipulation tasks. Our method achieves superior performance over existing image-based methods and approaches the effectiveness of point-cloud-driven methods, while maintaining strong scalability as the number of agents increases. Further ablation studies demonstrate Gau DP s robustness and scalability as the number of agents increases. Visualization results confirm its ability to integrate multi-agent observations into a high-quality 3D global representation that improves decision accuracy. |
| Researcher Affiliation | Collaboration | 1Sun Yat-sen University 2The University of Hong Kong 3Shanghai Jiao Tong University 4The Chinese University of Hong Kong, Shenzhen 5Shanghai AI Laboratory |
| Pseudocode | No | The paper describes the methods and framework using narrative text and diagrams (e.g., Figure 2) but does not include any explicit pseudocode or algorithm blocks. |
| Open Source Code | Yes | Codes are available at https://ziyeeee.github.io/gaudp.io/. |
| Open Datasets | Yes | To address this issue, we leverage the Robo Factory benchmark [1], an automated data collection framework specifically designed for embodied multi-agent systems. [1] Y. Qin, L. Kang, X. Song, Z. Yin, X. Liu, X. Liu, R. Zhang, and L. Bai, Robofactory: Exploring embodied agent collaboration with compositional constraints, ar Xiv preprint ar Xiv:2503.16408, 2025. |
| Dataset Splits | Yes | Evaluation is performed every 100 training epochs over 100 episodes per policy. |
| Hardware Specification | Yes | All experiments are implemented using the Py Torch framework and conducted on a single NVIDIA A800 GPU. Training on A100 GPU; inference on NVIDIA RTX 5090 GPU. |
| Software Dependencies | No | The paper states, 'All experiments are implemented using the Py Torch framework,' but it does not specify version numbers for PyTorch or any other software dependencies. |
| Experiment Setup | Yes | Policies are trained for 100 epochs using a batch size of 32. We adopt the Adam optimizer with an initial learning rate of 10 4, combined with a warm-up phase followed by cosine decay scheduling. Specifically, we use an action prediction horizon of 8, 3 observation steps, and 6 action execution steps. Both Gau DP and DP adopt DDPM with 100 denoising steps, while DP3 employs DDIM with the same number of steps. |