Reinforcement Learning Method Advances AI-Driven Crystal Design for Functional Materials

Zhendong Cao and Lei Wang of the Chinese Academy of Sciences present a reinforcement learning-based method that advances AI-driven crystal design. While generative machine learning has progressed crystal discovery, it struggles to find candidates that are both novel and useful. The new reinforcement learning loop steers generation to these optimal areas, enabling the design of novel functional materials.

Reinforcement learning for crystals

• ▪A reinforcement learning-based method steers candidate generation to areas of material candidates that are both novel and useful.
• ▪Zhendong Cao and Lei Wang of the Chinese Academy of Sciences published research on reinforcement learning steering generative crystal design on August 3, 2026.

Generative AI materials discovery

• ▪Generative machine learning methods cannot fully explore the space of material candidates that are both novel and useful.
• ▪Generative machine learning methods have led to progress in crystal discovery.

Goal-directed crystal generation

• ▪H. Park and A. Walsh published research in Nature Machine Intelligence in 2026 regarding generative crystal design.
• ▪A reinforcement learning loop is utilized to achieve goal-directed crystal generation.

Functional materials design

• ▪The reinforcement learning-based method enables the design of novel functional materials.
• ▪The design of novel functional materials is limited by the inability of traditional generative methods to find both novel and useful candidates.

1 source