The advent of machine learning (ML) has been a transformative force in the realm of computational solid mechanics, creating a platform for the development of novel methods with unprecedented efficiency. Nonetheless, several limitations still hinder their direct incorporation for industry-level engineering problems, including the need to generate vast datasets, their limited generalizability beyond the training regime, their numerical sensitivity to the choice of hyperparameters, and more. Thus, there is an increasing effort to develop new hybrid modeling paradigms where ML-based techniques are utilized in conjunction with conventional numerical approaches. The goal of such methods is to harness the strengths of both worlds, that is the robustness of well-established conventional approaches with the effectiveness of machine-learning applications, and ultimately take upon real-world, complex solid mechanics problems. This minisymposium will explore the latest developments across this front, and without being limited to the following, it invites contributions that delve into:

• Novel combinations of ML-methods with the Finite Element Method (FEM), Boundary Element Method (BEM), Virtual Element Method (VEM), meshless approaches, and more.
• Fusion of ML-tools into multi-scale (FE2) modeling, to accelerate solid mechanics simulations across different length scales.
• Failure mechanics analysis using physics-augmented machine learning, with applications on phase-field fracture, continuum damage, peridynamics, cohesive zone models, etc.
• Challenges and opportunities in the utilization of ML-methods for the solution of coupled multi-physics problems.
• Efficient and robust pathways to extrapolate the ML predictive capability beyond the bounds of the training region.
• Stochastic analysis and uncertainty quantification of such hybrid methods.