---
license: cc-by-nc-sa-4.0
tags:
- MLIP
- MLFF
datasets:
- aixsim/Titan25
---

# SNet-T25
SNet-T25 is an MLIP trained on [Titan25 dataset](https://huggingface.co/datasets/aixsim/Titan25).
It was designed for general chemical reactivity, including interfacial reactions,
and trained on 1.8M data points across 11 elements (C, H, N, O, Ag, Au, Cu, Pd, Pt, Rh, Ru).
The evaluation results are available in [GAIA paper](https://arxiv.org/abs/2509.25798),
which also provides full details of the model architecture and hyperparameters.

The original SNet-T25 was trained with our in-house framework;
we provide a converted checkpoint that can be used directly with [SevenNet framework](https://github.com/MDIL-SNU/SevenNet).
To enable compatibility, the user only need to apply the following minor modifications:
- Modify line 5 in `SevenNet/sevenn/__init__.py` to, e.g., `__version__ = '0.11.0'`. This allows the user to use it without installing SevenNet.
- Comment out line 332 in `SevenNet/sevenn/checkpoint.py`, i.e., `assert len(missing) == 0, f'Missing keys: {missing}'`

# Citation
If using this checkpoint, please cite our work as follows:
```
@article{gaia2025,
  title={Scalable Reactive Atomistic Dynamics with GAIA},
  author={Song, Suhwan and Kim, Heejae and Jang, Jaehee and Cho, Hyuntae and Kim, Gunhee and Kim, Geonu},
  journal={arXiv preprint arXiv:2509.25798},
  year={2025}
}
```