# PLaID++

This repository contains our flagship model's weights in our paper: [**_PLaID++: A Preference-Aligned Language Model for Targeted Inorganic Materials Design_**](https://arxiv.org/pdf/2509.07150), by [Andy Xu](https://www.linkedin.com/in/andyxuai/), [Rohan Desai](https://www.rohandesai.xyz), [Larry Wang](https://www.linkedin.com/in/larwang314/), [Gabriel Hope](https://www.linkedin.com/in/gabriel-hope-87472542/), and [Ethan Ritz](https://www.linkedin.com/in/ethan-ritz-2bba69382/).

## Summary

PLaID++ introduces an LLM fine-tuned for stable and property-targeted inorganic crystal generation. PLaID++ achieves a **~50% higher S.U.N.** (Stable, Unique, Novel) rate than prior work and robust conditional generation by space group though:
1. Leveraging a novel Wyckoff-based text encoding
2. Aligning the model using Direct Preference Optimization (DPO), an RL method guided by machine-learned interatomic potentials
3. Unified training across conditional and unconditional generation tasks

![plaid_architecture_diagram](https://arxiv.org/html/2509.07150v1/Figures/plaid++_diagram.png)

## Model

[The full PLaID++ model is available in `train_dpo/`](https://huggingface.co/HOPE-Lab-HMC/PLaID/tree/main/train_dpo).

## Citation

[Arxiv Link](https://arxiv.org/pdf/2509.07150)
```
@article{xu2025plaid++,
  title={PLaID++: A Preference Aligned Language Model for Targeted Inorganic Materials Design},
  author={Xu, Andy and Desai, Rohan and Wang, Larry and Hope, Gabriel and Ritz, Ethan},
  journal={arXiv preprint arXiv:2509.07150},
  year={2025}
}
```

## License

Most of PLaID++ is distributed under the CC BY 4.0 license. However, some components of the project are governed by different licenses: pymatgen is licensed under MIT, Hugging Face Transformers under Apache 2.0, and ASE under the GNU Lesser General Public License.