In-context learning as a retraining-free paradigm for constitutive modeling.

In-Context Modeling (ICM) enables retraining-free constitutive modeling by inferring hyperelastic stress responses from physical context, trained once on equilibrium fields spanning diverse materials.

• Why it matters: Constitutive modeling often needs material-specific retraining or recalibration when materials change, limiting reuse across new geometries and boundary conditions.
• What they did: ICM is trained once using equilibrium deformation fields across diverse materials, geometries, and boundary conditions, enforcing physical equilibrium without stress labels, then deployed on unseen hyperelastic materials using a single forward pass with nonuniform strain-field context.
• The result: Benchmarks and experimental validation show accurate stress prediction for out-of-distribution materials and boundary-value problems, with reduced deployment cost versus per-material learning, supporting ICM as a retraining-free paradigm.

AI summary of the abstract · may contain errors · not medical advice · read the paper.