I want to evaluate AI on my research
From reproducing a published result to running a simulation. You set the standard it's judged by.
Where Research Comes Together
OpenRefinery partners with scientists to advance AI for Science and expand the impact of their research.
We pursue joint research and publications, and create opportunities for training, career growth, and research funding.
You retain ownership and credit as we connect your research with industry R&D, helping businesses accelerate discovery through automated research and helping academic teams bring their work into practice.
We Work With Researchers At Stanford, MIT, Caltech, ETH Zürich, UPenn, Berkeley, NUS, And Other Institutions
We partner with researchers to evaluate AI on scientific work, from reproducing a result to running a simulation. Researchers define what a correct result requires, and our engineers help build the evaluations and benchmarks.
Researchers in our community
How It Works
Share A Research Problem
Share a question you are investigating, a method you use, or a result you cannot explain. This could be a calculation that does not match measurements, a result that changes with the conditions, or a method whose limits need closer study.
We Help You Build It
You define the problem and what a correct result requires. Our engineers help turn your methods and data into evaluations, benchmarks, or training datasets others can license.
You Keep Ownership
With your consent, we license these resources to AI labs on your behalf and share the revenue to support your continued research.
What Are You Working On?
Tell us about your research
I want to meet other researchers
Compare methods, discuss results, and get to know each other over coffee and hangouts. Starting in the Bay Area, with online gatherings planned.
I have a dataset others could use
We assess raw or prepared datasets, transform them for training and evaluation, and arrange licensing under agreed terms.
I need data for my research
Tell us the capability you want to evaluate or advance. We assess available datasets and explain their coverage, preparation needs, and limits.
