The Well: A Multi-Domain Benchmark for Physics Foundation Models
The Well paper systematically organizes data from 16 physics simulation domains and proposes a unified interface and evaluation protocol, arguing for the feasibility and benefit of cross-domain pretraining (Physics Foundation Model). It is a must-read for entering physics-AI foundation-model research.
Highlights
- Unified multi-domain interface lowers cross-domain training barriers
- Pretraining baselines enable fair comparison
- Emphasizes data quality and diversity for generalization
This entry can be auto-aggregated periodically from arXiv by the AI pipeline (M5).
Tags: BenchmarkPhysics Foundation ModelSurveyMust-Read
Related entries
- Physics-Informed Learning Annual Review 2025-2026 · A systematic review of 2025-2026 progress in physics-informed ML - neural operat
- PhysicsNeMo: Neural-Operator-Based Physics Simulation Framework · NVIDIA's PINN framework built on neural operators (FNO/DeepONet/AFNO) for end-to
- Transolver: A Transformer Solver for Large-Scale Industrial Geometries · Proposes a physics-attention mechanism that projects meshes onto learnable physi
- DeepONet (Original Paper) · Presents a branch-trunk architecture realizing a deep learning version of the un
- Fourier Neural Operator (Original Paper) · Introduces neural operators with integral kernels parameterized in Fourier space