AtomWorld: A Benchmark for Evaluating Spatial Reasoning in Large Language Models on Material Structures

Taoyuze Lv, Alexander Chen, Fengyu Xie, Chu Wu, Jeffrey Meng, Dongzhan Zhou, Bram Hoex, Yingheng Wang, Zhicheng Zhong, Tong Xie
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:83497-83523, 2026.

Abstract

Large language models (LLMs) have shown promising potential in materials science, enabling tasks ranging from knowledge retrieval to property prediction. Existing materials science benchmarks mainly focus on perceptual or knowledge-based tasks, largely ignoring the structure modelling tasks, a core challenge in real scientific workflows. In practice, constructing and manipulating atomic structures is one of the most creative and least automated steps in materials research. In this work, we introduce AtomWorld, a benchmark designed to evaluate the abilities of LLMs on structure modifications. The benchmark includes ten fundamental actions under four widely used modelling categories, enabling verifiable evaluation metrics. We find that Gemini 2.5 Pro generally performs the best. While the success rate decreases markedly with increasing modelling complexity, with particularly low success rates (below 12% for rotation) for operations involving complex spatial relations. Our results suggest that contemporary LLMs are better suited as copilots for materials structure modelling rather than fully unsupervised autonomous scientific agents. Beyond evaluation, AtomWorld also serves as a testbed and playground for developing future structure-aware models, including reinforcement learning and agentic approaches.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-lv26b, title = {{A}tom{W}orld: A Benchmark for Evaluating Spatial Reasoning in Large Language Models on Material Structures}, author = {Lv, Taoyuze and Chen, Alexander and Xie, Fengyu and Wu, Chu and Meng, Jeffrey and Zhou, Dongzhan and Hoex, Bram and Wang, Yingheng and Zhong, Zhicheng and Xie, Tong}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {83497--83523}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/lv26b/lv26b.pdf}, url = {https://proceedings.mlr.press/v306/lv26b.html}, abstract = {Large language models (LLMs) have shown promising potential in materials science, enabling tasks ranging from knowledge retrieval to property prediction. Existing materials science benchmarks mainly focus on perceptual or knowledge-based tasks, largely ignoring the structure modelling tasks, a core challenge in real scientific workflows. In practice, constructing and manipulating atomic structures is one of the most creative and least automated steps in materials research. In this work, we introduce AtomWorld, a benchmark designed to evaluate the abilities of LLMs on structure modifications. The benchmark includes ten fundamental actions under four widely used modelling categories, enabling verifiable evaluation metrics. We find that Gemini 2.5 Pro generally performs the best. While the success rate decreases markedly with increasing modelling complexity, with particularly low success rates (below 12% for rotation) for operations involving complex spatial relations. Our results suggest that contemporary LLMs are better suited as copilots for materials structure modelling rather than fully unsupervised autonomous scientific agents. Beyond evaluation, AtomWorld also serves as a testbed and playground for developing future structure-aware models, including reinforcement learning and agentic approaches.} }
Endnote
%0 Conference Paper %T AtomWorld: A Benchmark for Evaluating Spatial Reasoning in Large Language Models on Material Structures %A Taoyuze Lv %A Alexander Chen %A Fengyu Xie %A Chu Wu %A Jeffrey Meng %A Dongzhan Zhou %A Bram Hoex %A Yingheng Wang %A Zhicheng Zhong %A Tong Xie %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-lv26b %I PMLR %P 83497--83523 %U https://proceedings.mlr.press/v306/lv26b.html %V 306 %X Large language models (LLMs) have shown promising potential in materials science, enabling tasks ranging from knowledge retrieval to property prediction. Existing materials science benchmarks mainly focus on perceptual or knowledge-based tasks, largely ignoring the structure modelling tasks, a core challenge in real scientific workflows. In practice, constructing and manipulating atomic structures is one of the most creative and least automated steps in materials research. In this work, we introduce AtomWorld, a benchmark designed to evaluate the abilities of LLMs on structure modifications. The benchmark includes ten fundamental actions under four widely used modelling categories, enabling verifiable evaluation metrics. We find that Gemini 2.5 Pro generally performs the best. While the success rate decreases markedly with increasing modelling complexity, with particularly low success rates (below 12% for rotation) for operations involving complex spatial relations. Our results suggest that contemporary LLMs are better suited as copilots for materials structure modelling rather than fully unsupervised autonomous scientific agents. Beyond evaluation, AtomWorld also serves as a testbed and playground for developing future structure-aware models, including reinforcement learning and agentic approaches.
APA
Lv, T., Chen, A., Xie, F., Wu, C., Meng, J., Zhou, D., Hoex, B., Wang, Y., Zhong, Z. & Xie, T.. (2026). AtomWorld: A Benchmark for Evaluating Spatial Reasoning in Large Language Models on Material Structures. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:83497-83523 Available from https://proceedings.mlr.press/v306/lv26b.html.

Related Material