A single AI model can predict the behavior of a brand-new material it has never seen before in one second.
April 29, 2026
Original Paper
In-context modeling as a retrain-free paradigm for foundation models in computational science
arXiv · 2604.23098
The Takeaway
Scientists usually have to train a new model for every specific material or physical system they want to study. This retrain-free paradigm uses observational data as a prompt, allowing one model to understand a wide range of physical relationships instantly. The AI can infer how stress moves through a new geometry or how heat flows through a novel substance in a single forward pass. This treats the laws of physics as a context for the model rather than something that must be hard-coded. This method could drastically accelerate discovery in material science and engineering.
From the abstract
Building models that generalize across physical systems without retraining remains a central challenge in computational science. Here we introduce In-Context Modeling (ICM), a retrain-free paradigm that infers physical relationships directly from observational fields. Rather than encoding system-specific behavior in fixed parameters, ICM assimilates measurements as physical context and performs inference through a single forward pass. Trained in a physics-informed, label-free manner using govern