metadata
license: mit
pretty_name: SIMSHIFT data
SIMSHIFT: A Benchmark for UDA in Industrial Surrogates
SIMSHIFT is a benchmark suite of simulation datasets for Unsupervised Domain Adaptation and Neural Surogates for physical simulations. This is the official data repository to the NeurIPS 2025 Datasets & Benchmarks Track Submission.
We provide four simulation datasets, motivated by industrial relevance:
- Hot Rolling
- Sheet Metal Forming
- Electric Motor Design
- Heatsink Design
Usage
We provide dataset loading utilities and full training and evaluation pipelines in the official code repository.