Chem World: A Large-Scale Benchmark and Physics-Informed Framework for Trustworthy Chemical Property Prediction

Standardized benchmarks are essential for trustworthy AI: The paper addresses a critical gap in responsible AI development by creating Chem World, a unified evaluation platform combining 17 datasets with over 800,000 molecular samples. This standardization allows researchers to consistently assess whether AI models are reliable and generalizable—key requirements before deploying AI in high-stakes domains like drug development.

Physics-informed AI outperforms pure data-driven approaches: The proposed Mixture-PINN framework improves prediction accuracy, robustness, and reliability by incorporating established chemical knowledge into neural networks. This demonstrates that trustworthy AI systems benefit from combining domain expertise with machine learning, rather than relying solely on data patterns.

Task diversity matters for real-world reliability: By evaluating models across multiple property types (density, electrical conductivity, solubility, etc.) rather than single tasks, Chem World reveals whether AI systems truly generalize or merely memorize specific patterns. This multi-task approach better reflects how AI will actually be used in chemistry and drug discovery.

Fragmented evaluation undermines responsible AI development: The paper identifies that existing benchmarks suffer from "limited task diversity, fragmented datasets, and inconsistent evaluation protocols"—problems that make it impossible to reliably assess whether an AI model is trustworthy before deployment in scientific research or drug development.