Search papers, labs, and topics across Lattice.
This paper benchmarks the out-of-distribution transfer performance of PDE foundation models (POSEIDON and MORPH) on extreme-loading material dynamics tasks involving shocks and fractures. The downstream task is formulated as predicting the terminal state from the initial state, and the study compares fine-tuning from pretrained weights against training from scratch. Results quantify the sample efficiency of pretrained models under distribution shift in two discontinuity-dominated regimes: shock-driven multi-material interface dynamics and dynamic fracture/failure evolution.
PDE foundation models trained on fluid dynamics struggle to generalize to material dynamics under extreme loading, highlighting a significant gap in their out-of-distribution capabilities.
Most PDE foundation models are pretrained and fine-tuned on fluid-centric benchmarks. Their utility under extreme-loading material dynamics remains unclear. We benchmark out-of-distribution transfer on two discontinuity-dominated regimes in which shocks, evolving interfaces, and fracture produce highly non-smooth fields: shock-driven multi-material interface dynamics (perturbed layered interface or PLI) and dynamic fracture/failure evolution (FRAC). We formulate the downstream task as terminal-state prediction, i.e., learning a long-horizon map that predicts the final state directly from the first snapshot without intermediate supervision. Using a unified training and evaluation protocol, we evaluate two open-source pretrained PDE foundation models, POSEIDON and MORPH, and compare fine-tuning from pretrained weights against training from scratch across training-set sizes to quantify sample efficiency under distribution shift.
