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..
DiffPhyCon: A Generative Approach to Control Complex Physical Systems
Authors: Long Wei, Peiyan Hu, Ruiqi Feng, Haodong Feng, Yixuan Du, Tao Zhang, Rui Wang, Yue Wang, Zhi-Ming Ma, Tailin Wu
NeurIPS 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We test our method on three tasks: 1D Burgers equation, 2D jellyfish movement control, and 2D highdimensional smoke control, where our generated jellyfish dataset is released as a benchmark for complex physical system control research. Our method outperforms widely applied classical approaches and state-of-the-art deep learning and reinforcement learning methods. |
| Researcher Affiliation | Collaboration | 1School of Engineering, Westlake University, 2Academy of Mathematics and Systems Science, Chinese Academy of Sciences, 3Jilin University, 4Fudan University, 5Microsoft AI4Science |
| Pseudocode | Yes | Algorithm 1 Inference for Diff Phy Con |
| Open Source Code | Yes | The project website, jellyfish dataset, and code can be found at https://github.com/AI4Science-Westlake U/diffphycon. |
| Open Datasets | Yes | our generated jellyfish dataset is released as a benchmark for complex physical system control research. |
| Dataset Splits | No | The paper specifies 'training set' and 'testing set' sizes for all experiments, but does not explicitly mention or detail a 'validation set' or its use for hyperparameter tuning. |
| Hardware Specification | Yes | The training is performed on two NVIDIA Tesla A100 GPUs with 80 GB memory for about 3 days. ... The training is performed on two NVIDIA Tesla A6000 GPUs with 48 GB memory for about 2 days. ... Inference time is tested on a Tesla-V100 GPU with 8 CPUs. |
| Software Dependencies | No | The paper mentions software tools like 'Lily-Pad simulator' and 'Phiflow solver' and optimizers like 'Adam' but does not provide specific version numbers for these or any other software dependencies, such as deep learning frameworks. |
| Experiment Setup | Yes | Table 5: Hyperparameters of the UNet architecture and training for the results of 1D Burgers equation in Table 1. |