Code-Generated Graph Representations Using Multiple LLM Agents for Material Properties Prediction

Jiao Huang, Qianli Xing, Jinglong Ji, Bo Yang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:25972-25986, 2025.

Abstract

Graph neural networks have recently demonstrated remarkable performance in predicting material properties. Crystalline material data is manually encoded into graph representations. Existing methods incorporate different attributes into constructing representations to satisfy the constraints arising from symmetries of material structure. However, existing methods for obtaining graph representations are specific to certain constraints, which are ineffective when facing new constraints. In this work, we propose a code generation framework with multiple large language model agents to obtain representations named Rep-CodeGen with three iterative stages simulating an evolutionary algorithm. To the best of our knowledge, Rep-CodeGen is the first framework for automatically generating code to obtain representations that can be used when facing new constraints. Furthermore, a type of representation from generated codes by our framework satisfies six constraints, with codes satisfying three constraints as bases. Extensive experiments on two real-world material datasets show that a property prediction method based on such a graph representation achieves state-of-the-art performance in material property prediction tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-huang25an, title = {Code-Generated Graph Representations Using Multiple {LLM} Agents for Material Properties Prediction}, author = {Huang, Jiao and Xing, Qianli and Ji, Jinglong and Yang, Bo}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {25972--25986}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/huang25an/huang25an.pdf}, url = {https://proceedings.mlr.press/v267/huang25an.html}, abstract = {Graph neural networks have recently demonstrated remarkable performance in predicting material properties. Crystalline material data is manually encoded into graph representations. Existing methods incorporate different attributes into constructing representations to satisfy the constraints arising from symmetries of material structure. However, existing methods for obtaining graph representations are specific to certain constraints, which are ineffective when facing new constraints. In this work, we propose a code generation framework with multiple large language model agents to obtain representations named Rep-CodeGen with three iterative stages simulating an evolutionary algorithm. To the best of our knowledge, Rep-CodeGen is the first framework for automatically generating code to obtain representations that can be used when facing new constraints. Furthermore, a type of representation from generated codes by our framework satisfies six constraints, with codes satisfying three constraints as bases. Extensive experiments on two real-world material datasets show that a property prediction method based on such a graph representation achieves state-of-the-art performance in material property prediction tasks.} }
Endnote
%0 Conference Paper %T Code-Generated Graph Representations Using Multiple LLM Agents for Material Properties Prediction %A Jiao Huang %A Qianli Xing %A Jinglong Ji %A Bo Yang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-huang25an %I PMLR %P 25972--25986 %U https://proceedings.mlr.press/v267/huang25an.html %V 267 %X Graph neural networks have recently demonstrated remarkable performance in predicting material properties. Crystalline material data is manually encoded into graph representations. Existing methods incorporate different attributes into constructing representations to satisfy the constraints arising from symmetries of material structure. However, existing methods for obtaining graph representations are specific to certain constraints, which are ineffective when facing new constraints. In this work, we propose a code generation framework with multiple large language model agents to obtain representations named Rep-CodeGen with three iterative stages simulating an evolutionary algorithm. To the best of our knowledge, Rep-CodeGen is the first framework for automatically generating code to obtain representations that can be used when facing new constraints. Furthermore, a type of representation from generated codes by our framework satisfies six constraints, with codes satisfying three constraints as bases. Extensive experiments on two real-world material datasets show that a property prediction method based on such a graph representation achieves state-of-the-art performance in material property prediction tasks.
APA
Huang, J., Xing, Q., Ji, J. & Yang, B.. (2025). Code-Generated Graph Representations Using Multiple LLM Agents for Material Properties Prediction. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:25972-25986 Available from https://proceedings.mlr.press/v267/huang25an.html.

Related Material