Graph-based models provide the lowest errors in property prediction, retain their advantage across the evaluated training-set sizes, and remain robust to increasing repeat-unit complexity, according to the PolyBench26 benchmark.
Abstract
Polymer property prediction lacks open, standardized benchmarks that enable rigorous comparison of machine-learning methods, with existing resources covering only a narrow fraction of polymer architectures, such as homopolymers. We introduce Polymer Benchmark 2026 (PolyBench26), an open dataset comprising nearly 250,000 polymer-property datapoints across eight physical properties, including data from experimental measurements, density functional theory, and molecular dynamics. The benchmark supports four evaluation tasks across homopolymers and alternating, random, and block copolymers: in-distribution property prediction, dataset-size scaling, repeat-unit complexity, and transfer to held-out polymer architectures. We compare language model, graph-based, and descriptor-based approaches and find graph-based models provide the lowest errors in property prediction, retain their advantage across the evaluated training-set sizes, and remain robust to increasing repeat-unit complexity. PolyBench26 provides a reproducible foundation for developing models for the increasingly complex polymer design space. The PolyBench26 benchmark is available open-source at https://github.com/rlearsch/PolymerBenchmark2026.
PolyBench is a public benchmark platform comprising 39 literature-derived prediction tasks with 42,169 curated structure–property records covering 14,717 distinct repeat-unit structures, seven repeat-unit molecular representation types, five fingerprint families, physicochemical descriptors, and task-trained graph embe...
Polymers are fundamental to modern materials science because their backbone chemistry, monomer composition, and chain architecture can be systematically tuned to achieve a virtually unlimited range of mechanical, thermal, electronic, and optical properties. To accelerate the discovery and design of polymeric material...
Amberbir Alemayoh, Zhi-Xin Pan, Ning Wang· Polymer Science & Techno...· 0 citations
A data-driven polymer informatics framework is developed by integrating molecular dynamics simulations with machine learning models to enable rapid and accurate prediction of key structural descriptors, including the radius of gyration and solvent-accessible surface area.
Lorand Gabriel Parajdi, I. Tóth, Alex-Adrian Farcas et al.· Scientific Reports· 0 citations
Uni-Macro-FRPN (FRPN), a Full-Resolution Polymer Network that retains both detailed atom-level and monomer-level features and explicit polymer structure information within a unified framework, is presented, providing a practical route for moving polymer representation learning beyond monomer-centric representations.
This study explores data-driven approaches for predicting the thermal decomposition temperature of polymers using both classical machine learning (CML) models and a small language model (SLM), suggesting that small language models can serve as a valuable alternative modeling strategy for predicting polymer thermal prop...
Nguyen T. T. Duyen, Ngo T. Que, Hanh Bich Vu· Journal of Physics, Conferen...· 0 citations
An interpretable machine learning workflow is developed that predicts both the total dielectric constant and the HSE bandgap of polymer repeat units directly from a monomer structure, using an open density-functional-theory dataset of 284 four-block polymers.
F. Gül· Polymers· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduOct 13, 2026
Machine-learning networks can learn without detailed instructions from a designer. Those that do may offer insight into how the brain reorganizes itself during learning.
MIT News · Artificial Intelligence· news.mit.eduSep 29, 2026
Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.