NeurIPS 2026 Workshop
XAI4Science
Knowledge Discovery and Trust through Interpretable Foundation Models
Dec 11–12, 2026
Sydney, Australia
Co-located with NeurIPS 2026
Weather and climate foundation models now rival physics-based forecasts at a fraction of the computational cost. But unlike models built on known equations, their reasoning is opaque, and they can quietly hallucinate. XAI4Science brings together ML researchers, climate scientists, and policy experts to make these models explainable, trustworthy, and physically grounded.
What the Workshop Covers
01
Ante-hoc
Ante-hoc
Interpretability
Self-explainable architectures and inductive biases for weather and climate FMs.
02
Post-hoc Attribution & Evaluation
Attribution and probing methods, evaluated for whether they're actually faithful.
03
Physics-Consistent Explanations
Benchmarks that check explanations against known physical laws.
This is not an exhaustive list and closely related topics are welcome. If you have questions about the fit do not hesitate to reach out xai4science@gmail.com.
Organizers
Leonardo Pesce* · Jiawen Wei* · Max Welling** · Gianmarco Mengaldo*
*National University of Singapore **University of Amsterdam
XAI4Science is non-profit and community-run.
Questions or sponsorship inquiries welcome.