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..
Environment Inference for Learning Generalizable Dynamical System
Authors: Shixuan Liu, Yue He, Haotian Wang, Wenjing Yang, Yunfei Wang, Peng Cui, Zhong Liu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments investigate three dynamical systems governed by specific differential equations: an ODE for biological modeling, PDEs for reaction-diffusion in chemistry, and the Navier-Stokes equations for incompressible fluid dynamics. These complex, nonlinear systems test our method s ability to classify spatio-temporal patterns and physical laws across diverse environments. ... The in-domain generalization results detailed in Table 1 illustrate the performance implications of various assignment strategies. |
| Researcher Affiliation | Academia | 1College of Computer Science and Technology, National University of Defense Technology 2School of Information, Renmin University of China 3College of Systems Engineering, National University of Defense Technology 4Department of Computer Science and Technology, Tsinghua University 5Laboratory for Big Data and Decision, National University of Defense Technology |
| Pseudocode | Yes | Algorithm 1 Dyna Infer framework |
| Open Source Code | Yes | Code is available at https://github.com/shixuanliu-andy/Dyna Infer |
| Open Datasets | Yes | Lotka Volterra (LV) [23] The system models the dynamics between a prey-predator pair in an ecosystem, captured by the following ODE: ... Gray-Scott (GS) [28] The model uses simple reaction-diffusion equations to effectively study complex pattern formation in chemical and biological systems, following underlying PDE dynamics: ... Navier-Stokes (NS) [22] The Navier-Stokes PDE describes the motion of viscous fluid substances: ... |
| Dataset Splits | Yes | For in-domain experiments, we generate four LV trajectories in each of nine environments, ten GS trajectories in each of three environments, and eight NS trajectories in each of four environments. For adaptation experiments, we simulate the same number of trajectories per environment, conducting finetuning in two additional environments e Eu. All dynamic environment parameters are detailed in Appendix E. For evaluation, we sample 32 trajectories per environment, initialized according to the underlying distribution p(x0). The LV and GS data are generated using the DOPRI5 solver [8, 12], while the NS data is simulated with the pseudo-spectral method as in [22]. ... The dataset was partitioned into training and test sets with a ratio of 6:1. Furthermore, each trajectory was temporally resampled to a standardized length of 100 time steps for smoothing and uniformity following the practice in [44]. |
| Hardware Specification | Yes | We conducted experiments on a server equipped with a 64-core CPU, 256 GB of RAM, and eight 24GB RTX-3090Ti GPUs. |
| Software Dependencies | No | The Dyna Infer framework was implemented using Py Torch [27]. |
| Experiment Setup | No | All neural network architectures, optimizers, and parameters for the base models are configured as described in their respective papers. |