Marius Tacke
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Marius Tacke has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Christian Johannes Cyron
Hamburg University of Technology
4 shared publications - Matthias Busch
Institute for Artificial Intelligence and Simulation in Mechanics Hamburg University of Technology Hamburg Germany
3 shared publications - Roland C. Aydin
Institute for Continuum and Material Mechanics, Hamburg University of Technology, Hamburg, Germany
3 shared publications - Christian Feiler
Helmholtz-Zentrum Hereon
2 shared publications Kevin Linka
Institute of Applied Medical Engineering RWTH Aachen University Aachen Germany
2 shared publications- Mikhail L. Zheludkevich
Institute of Surface Science Helmholtz‐Zentrum Hereon Geesthacht Germany
2 shared publications
Latest Publications
- Open Access
- 19 October 2026
- Published by Wiley in GAMM-Mitteilungen
- Vol. 49 (4) , e70016
Agentic artificial intelligence in mechanics of materials: A review of building tailored models on demand
In mechanics of materials, many powerful methods are accessible only to experts. Constitutive models, for example, describe the relation between stress and strain in materials and are a core ingredient of simulations from structural engineering to biomechanics. Yet building and using such models requires specialized expertise—a barrier for the many potential users trained in other fields. Artificial intelligence based on foundation models, increasingly organized as agents, is beginning to lower this barrier by allowing users to generate tailored models on demand. The same development is underway in many other fields where specialized tools have long been hand‐crafted anew for each task. We call this emerging paradigm the on‐the‐fly generation of physical models : from a plain request, an agent creates a physical model on demand—from scratch and tailored to the problem. Its appeal is accessibility and flexibility. As a consequence, the lasting artifact is no longer a fixed model but the orchestration and prompting of the agents that produce it. This review surveys the concepts behind the paradigm, within mechanics of materials and in neighboring fields, and examines them through two lenses: effectiveness (the accuracy of the generated models) and trustworthiness (their physical admissibility). Trustworthiness is pursued along two routes: deepening the agent's integration under explicit validation, or restraining its autonomy in favor of trusted components. Given the rapid progress of foundation models, and the bitter lesson that general methods leveraging computation tend to prevail, we call for paradigms designed to scale with this progress rather than to compete with it.
- 19 October 2026
- Published by Wiley in GAMM-Mitteilungen
- Vol. 49 (4) , e70016
Open AccessAgentic artificial intelligence in mechanics of materials: A review of building tailored models on demand
In mechanics of materials, many powerful methods are accessible only to experts. Constitutive models, for example, describe the relation between stress and strain in materials and are a core ingredient of simulations from structural engineering to biomechanics. Yet building and using such models requires specialized expertise—a barrier for the many potential users trained in other fields. Artificial intelligence based on foundation models, increasingly organized as agents, is beginning to lower this barrier by allowing users to generate tailored models on demand. The same development is underway in many other fields where specialized tools have long been hand‐crafted anew for each task. We call this emerging paradigm the on‐the‐fly generation of physical models : from a plain request, an agent creates a physical model on demand—from scratch and tailored to the problem. Its appeal is accessibility and flexibility. As a consequence, the lasting artifact is no longer a fixed model but the orchestration and prompting of the agents that produce it. This review surveys the concepts behind the paradigm, within mechanics of materials and in neighboring fields, and examines them through two lenses: effectiveness (the accuracy of the generated models) and trustworthiness (their physical admissibility). Trustworthiness is pursued along two routes: deepening the agent's integration under explicit validation, or restraining its autonomy in favor of trusted components. Given the rapid progress of foundation models, and the bitter lesson that general methods leveraging computation tend to prevail, we call for paradigms designed to scale with this progress rather than to compete with it.
- Open Access
- 1 October 2026
- Published by Springer Nature in Machine Learning for Computational Science and Engineering
- Vol. 2 (2) , 51
LLM-driven design of physics-constrained constitutive models: Two agents are better than one
Developing constitutive models that capture how materials deform under load traditionally requires years of specialized expertise in continuum mechanics, machine learning, and scientific programming. Large language models (LLMs) have recently been shown to lower this barrier by generating constitutive models on demand, but existing single-agent pipelines lack systematic checks that the resulting models respect fundamental physical laws. To close this gap, we introduce the first multi-agent LLM-driven approach for constitutive model generation: a Creator agent proposes a model tailored to the data, while an Inspector agent critically audits each proposal against nine physical constraints and returns it for refinement whenever a violation is detected. We demonstrate this concept with constitutive artificial neural networks (CANNs) and benchmark it on brain tissue and rubber as isotropic materials, and on porcine skin tissue as a transversely isotropic material with a preferred fiber direction, using two different LLM backbones (Claude Opus 4.7 and Kimi K2.5). Whether a generated model satisfies the physical constraints is assessed numerically, by probing each constraint across a broad sample of deformation states, rotations, and perturbation directions. Adding the Inspector raises the share of exported models that pass all these checks from 90% to 95% for Opus and from 47% to 60% for Kimi. In addition, the generated models are on par with or even surpass expert-designed models in accuracy, extrapolate reliably beyond the training data, and generalize remarkably well to unseen loading paths. Separating generation from inspection thus turns LLM-driven constitutive modeling into a substantially more trustworthy process. The paradigm is deliberately technique-agnostic and scales automatically with advances in LLM capability, opening a promising path toward automated, physics-aware model discovery.
- 1 October 2026
- Published by Springer Nature in Machine Learning for Computational Science and Engineering
- Vol. 2 (2) , 51
Open AccessLLM-driven design of physics-constrained constitutive models: Two agents are better than one
Developing constitutive models that capture how materials deform under load traditionally requires years of specialized expertise in continuum mechanics, machine learning, and scientific programming. Large language models (LLMs) have recently been shown to lower this barrier by generating constitutive models on demand, but existing single-agent pipelines lack systematic checks that the resulting models respect fundamental physical laws. To close this gap, we introduce the first multi-agent LLM-driven approach for constitutive model generation: a Creator agent proposes a model tailored to the data, while an Inspector agent critically audits each proposal against nine physical constraints and returns it for refinement whenever a violation is detected. We demonstrate this concept with constitutive artificial neural networks (CANNs) and benchmark it on brain tissue and rubber as isotropic materials, and on porcine skin tissue as a transversely isotropic material with a preferred fiber direction, using two different LLM backbones (Claude Opus 4.7 and Kimi K2.5). Whether a generated model satisfies the physical constraints is assessed numerically, by probing each constraint across a broad sample of deformation states, rotations, and perturbation directions. Adding the Inspector raises the share of exported models that pass all these checks from 90% to 95% for Opus and from 47% to 60% for Kimi. In addition, the generated models are on par with or even surpass expert-designed models in accuracy, extrapolate reliably beyond the training data, and generalize remarkably well to unseen loading paths. Separating generation from inspection thus turns LLM-driven constitutive modeling into a substantially more trustworthy process. The paradigm is deliberately technique-agnostic and scales automatically with advances in LLM capability, opening a promising path toward automated, physics-aware model discovery.
- Open Access
- 26 September 2026
- Published by Elsevier in Corrosion Science
- Vol. 273
Different alloy, Different story? Impact of transfer learning on predictive models for in silico screening of Mg dissolution modulators
- 26 September 2026
- Published by Elsevier in Corrosion Science
- Vol. 273
Open AccessDifferent alloy, Different story? Impact of transfer learning on predictive models for in silico screening of Mg dissolution modulators
- Open Access
- 20 June 2025
- Published by Elsevier in Corrosion Science
- Vol. 255
Large language models predicting the corrosion inhibition efficiency of magnesium dissolution modulators
- 20 June 2025
- Published by Elsevier in Corrosion Science
- Vol. 255
Open AccessLarge language models predicting the corrosion inhibition efficiency of magnesium dissolution modulators