Original title: HiPoly: a hierarchical polymer-native AI framework for property prediction and generative design
Ge Sun, Gervasio Zaldivar, Yuan Tian, Gustavo Perez Lemus +4 · September 2026
Polymers are large molecules made of repeating chemical units, and they are vital for energy, health, and transportation. However, artificial intelligence models struggle to capture their complex, multi-scale shapes. To solve this, researchers introduced HiPoly, a new AI system built specifically for polymers. HiPoly works by turning full polymer descriptions into a three-level hierarchical graph based on the G2RINSA representation used to describe polymer structures, capturing how monomers are linked. representation. This structure directly maps how units link together randomly, their chemical mixture, and their overall molecular weight. The system creates a full pipeline that moves from experimental recipe data to predicting material traits, designing new molecules, and testing them with physics simulations. The team found that HiPoly predicts heat and physical properties of complex polymer blends with state-of-the-art accuracy. They also used it to successfully design and confirm eco-friendly, PFASA group of very stable chemicals that do not break down in the environment, used in non-stick coatings and waterproofing.-free alternatives to long-lasting fluorinated plastics.
Imagine describing a forest to someone. If you list every tree, the description is too detailed. If you just call it a forest, you miss the variety. The trick is to describe trees, clusters of trees, and the whole forest. Here the same problem happens with complex materials: they have structure at the level of individual building blocks, how those connect into chains, and how chains form the bulk. HiPoly uses three levels in its representation to capture all scales at once.
This matters because it speeds up the search for new polymers—like finding safer alternatives to PFAS chemicals used in waterproof coatings. The framework can predict properties from experimental data and even design new molecules, then validate them with simulations. That means fewer lab experiments and faster progress on sustainable materials.
The abstract only demonstrates prediction accuracy for thermophysical properties; it does not claim success for other property types. The generative design example identifies candidates but has not been experimentally synthesized or tested outside of molecular simulations. The framework requires formulation data, which may not be available for all polymer systems.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
HiPoly: a hierarchical polymer-native AI framework for property prediction and generative design — Ge Sun, Gervasio Zaldivar, Yuan Tian et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.