2025 · ICML 2025 — MAS: Multi-Agent Systems in the Era of Foundation Models workshop · Preprint

xChemAgents: Agentic AI for Explainable Quantum Chemistry

Can Polat , Mehmet Tunçel , Mustafa Kurban , Erchin Serpedin , Hasan Kurban

Research highlights

xChemAgents: selecting and checking chemical descriptors

Use cooperating agents to choose descriptors and check physical consistency in quantum-property prediction.

Why we investigated it

Appending many descriptors can reduce accuracy and make a model harder to interpret.

What the work contributes

A Selector chooses a sparse weighted descriptor set, while a Validator checks units and scaling laws through dialogue.

What the evidence shows

The paper reports up to 22% lower mean absolute error than its compared baselines on the evaluated benchmarks.

Build on this work

Use the agent roles and validation workflow to investigate descriptor selection and interpretable property models.

  • Agentic chemistry
  • Explainable AI
  • Descriptor selection
  • Physical validation
Abstract
Recent progress in multimodal graph neural networks has demonstrated that augmenting atomic XYZ geometries with textual chemical descriptors can enhance predictive accuracy across a range of electronic and thermodynamic properties. However, naively appending large sets of heterogeneous descriptors often degrades performance on tasks sensitive to molecular shape or symmetry, and undermines interpretability. xChemAgents proposes a cooperative agent framework that injects physics-aware reasoning into multimodal property prediction. xChemAgents comprises two language-model-based agents: a Selector, which adaptively identifies a sparse, weighted subset of descriptors relevant to each target, and provides a natural language rationale; and a Validator, which enforces physical constraints such as unit consistency and scaling laws through iterative dialogue. On standard benchmark datasets, xChemAgents achieves up to a 22% reduction in mean absolute error over the state-of-the-art baselines, while producing faithful, human-interpretable explanations. Experiment results highlight the potential of cooperative, self-verifying agents to enhance both accuracy and transparency in foundation-model-driven materials science. The implementation and accompanying dataset are available at https://github.com/KurbanIntelligenceLab/xChemAgents.
Cite
@misc{polat2025xchemagents,
  title = {xChemAgents: Agentic AI for Explainable Quantum Chemistry},
  author = {Can Polat and Mehmet Tunçel and Mustafa Kurban and Erchin Serpedin and Hasan Kurban},
  year = 2025,
  journal = {ICML 2025 — MAS: Multi-Agent Systems in the Era of Foundation Models workshop},
  doi = {10.48550/arxiv.2505.20574},
  url = {https://arxiv.org/pdf/2505.20574},
}