What makes a useful prior?
Inverse problems often require knowledge beyond the measurements. We investigate how that knowledge can be represented, learned, and reused—and how it shapes the reliability of a solution.
toward
foundation models
for inverse problems
OpenInverse is a research initiative advancing AI for inverse problems across science and engineering. We study how models, learned priors, and evaluation methods can transfer across problems.
problem
An inverse problem has the same form wherever it appears.
y = A(x) + ηmeasurement · operator · unknown · noise
Inverse problems start with observations and ask what produced them: reconstructing an image from measurements, estimating a material’s properties, or recovering the structure beneath the Earth. The challenge is that incomplete and noisy observations can support more than one answer.
What changes between problems is the operator and the space the unknown lives in, while the difficulty stays the same. This is why a method developed for one problem keeps reappearing, in slightly different form, in problems that share nothing else.
question
How far can learning generalize across inverse problems?
There is more than one way for a method to generalize. It may be reusable within a single problem, portable across measurement setups, or transferable between problems that share no physics, and each of those claims requires different evidence.
S1
task-specific models
One problem, one setup, one dataset.
S2
reusable models
The same prior carries to a new instance.
S3
across operators and domains
Adapts to different forward operators and scientific domains.
S4
foundation models
Pretrained models that adapt across operators, tasks, and domains.
research
Exploring what learning can bring to inverse problems.
How can learning and domain knowledge work together?
We explore how AI and deep learning can combine with physics, mathematical models, numerical methods, and domain knowledge to solve inverse problems across science and engineering.
How far can a model generalize?
We study what can be shared across tasks, operators, and domains, and how to evaluate the capabilities and limitations of reusable and pretrained models.
benchmarks
OpenInverse Bench
OpenInverse Bench brings together problem definitions, datasets, baselines, and evaluation protocols to study how inverse methods perform—and how well they transfer.
Problem definitions
Operators, noise settings and splits, published as code.
Baselines
Classical and learned reference implementations, kept with the data.
Metrics
Reconstruction quality, measurement consistency, uncertainty, and computational cost.
Cross-problem evaluation
Trained on one operator family, tested on an unseen one.
programs
Research takes more than one group.
We organize challenges, competitions, workshops, and fellowships around shared inverse problems, connecting method researchers with domain expertise.
- challenges
One stated problem, a held-out set, a deadline.
- competitions
Leaderboard rounds on the benchmark.
- workshops
Method work and domain work in one room.
- fellowships
Longer work on a single open question.
contact
Tell us what you are trying to recover,
and what you can measure.
Working on an inverse problem? We’d like to hear about your research, your measurements, or a question you want to explore together.
contact@openinverse.com