Datasets:

scaling-group
/

icm-hyperelastic

license: apache-2.0
pretty_name: ICM Hyperelastic Dataset
task_categories:
  - tabular-regression
tags:
  - hyperelasticity
  - mechanics
  - plane-strain
  - plane-stress
  - finite-element

ICM Hyperelastic Dataset

This dataset accompanies the paper:

https://arxiv.org/abs/2604.23098

The corresponding code is available at:

https://github.com/scaling-group/icm-hyperelastic

It contains compressed dataset archives used for training and evaluating models for hyperelastic material problems.

Because the data are provided as compressed .tar.zst archives, the Hugging Face Dataset Viewer is not available for this dataset.

Dataset Splits

The archive names indicate the split and the mechanical setting:

Name pattern	Meaning
`train`	Training set
`val105`	Test-ID
`val11`	Test-M
`val13D`	Test-MGL and Test-MGL+
`pe`	Plane strain
`ps`	Plane stress

For example:

dataset-train-pe.tar.zst: training set under plane strain
dataset-train-ps.tar.zst: training set under plane stress
dataset-val105-pe.tar.zst: Test-ID under plane strain
dataset-val11-pe.tar.zst: Test-M under plane strain
dataset-val13D-pe.tar.zst: Test-MGL and Test-MGL+ under plane strain

Usage Notes

Except for the experimental validation results, which use the plane-stress dataset for training, all other results in the paper use the plane-strain dataset.

In the file names:

pe denotes plane strain.
ps denotes plane stress.

Citation

If you use this dataset, please cite the associated paper:

@article{li2026incontext,
  title = {In-context modeling as a retrain-free paradigm for foundation models in computational science},
  author = {Li, Lingfeng and Li, Zhuoyuan and Li, Shun and Zhan, Kaixin and Gao, Huajian and Chen, Changqing and Yang, Liu},
  year = {2026},
  url = {https://arxiv.org/abs/2604.23098},
  eprint = {2604.23098},
  archivePrefix = {arXiv},
  primaryClass = {cs.CE}
}