Sim2Science: ML with Imperfect Scientific Models

About the Workshop

AI4Science has matured into an established field, with ML now embedded throughout the simulator-based workflows of the natural sciences. Much of this progress runs through simulators—mechanistic models hand-crafted by domain experts and fit to data—that encode our scientific theories and underpin prediction, parameter inference, experimentation, and decision-making. Yet an ML method coupled to a simulator is only as good as that simulator: simulators simplify complex systems, omit intractable physics, and depend on uncertain parameters, creating a discrepancy between simulated and observed data that biases the scientific conclusions we draw.

The central question of this workshop is: How can we best leverage imperfect scientific simulators when confronted with real-world data, and how can ML help to account for and mitigate limitations in simulator-based workflows across a wide range of domains? Sim2Science is deliberately cross-domain: rather than focusing on a single scientific field, we bring together researchers who each maintain hierarchies of simulators at different fidelities—in chemistry, fusion, neuroscience, climate, and beyond—to build a shared vocabulary and toolkit for handling imperfect simulators, so that progress in one field can transfer to others.

Topics of Interest

We welcome contributions of any kind — new methods, applications, analyses, benchmarks, or position pieces — spanning biology, chemistry, physics, materials science, climate science, and related fields, as long as the work engages both machine learning and scientific simulators. Topics include:

• Simulation-based inference and related parameter inference methods
• Understanding and mitigating model misspecification, including simulator diagnostics and discrepancy modeling
• Emulator and surrogate modeling, as well as hybrid and physics-informed approaches
• Analysis of simulator structure, degeneracy, simplifications, and identifiability
• Simulator pipelines, including data handling, preprocessing, and integration with downstream ML models
• Active learning and Bayesian optimization for fitting parameters or model components
• Closed-loop and experiment-in-the-loop scientific workflows
• Multi-fidelity and multi-resolution modeling
• (Agentic) model and equation discovery
• Differentiable frameworks, LLM-assisted scientific reasoning, and workflow automation

Full submission tracks, instructions, and example simulators are on the Call for Papers page.

Sponsors

We gratefully acknowledge confirmed sponsorship from:

Contact

For questions or inquiries about the workshop, please contact us at:
sim2science@gmail.com