PDEBench and RealPDEBench
These are complementary benchmarks that address different scopes: PDEBench focuses on diverse synthetic PDE problems for ML model evaluation, while RealPDEBench extends this paradigm by introducing paired real-world and simulated data for validating ML methods on complex physical systems, making them useful together for comprehensive evaluation from synthetic to realistic settings.
About PDEBench
pdebench/PDEBench
PDEBench: An Extensive Benchmark for Scientific Machine Learning
Encompasses diverse PDE problems (diffusion-reaction, Navier-Stokes, shallow-water, Darcy flow, and others) with forward and inverse problem variants, alongside dataset generation tools using JAX and PyTorch. Provides pre-trained baseline models and integrates with DeepXDE, Hydra configuration management, and HDF5-based datasets hosted on Dataverse for reproducibility and extensibility.
About RealPDEBench
AI4Science-WestlakeU/RealPDEBench
[ICLR26 Oral] RealPDEBench: A Benchmark for Complex Physical Systems with Paired Real-World and Simulated Data
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work