International Verification of Neural Networks Competition (VNN-COMP)
Website: https://vnn-comp.github.io/
- Discussion: https://github.com/VNN-COMP/vnncomp2026
- Example benchmark: https://github.com/VNN-COMP/example_benchmark
- Example toolkit: https://github.com/VNN-COMP/example_toolkit
➡️ submit at https://vnn.repeatability.cps.cit.tum.de/
We have open-sourced the submission platform behind VNN-COMP.
This allows you to run the platform on a local server of your lab, and thoroughly test your submission before submitting to the real system.*
➡️ check it out at https://github.com/VNN-COMP/vnn-eval-platform
* Note: VNN-COMP runs on AWS instances, which cost quite a bit if all tool authors want to fully test their submissions, troubleshoot, and eventually run the full evaluation. Running it locally drastically reduces this cost. We assume here that most labs have some server setup where they could spin up a local version of the submission site; let us know if this is not the case. Please also note that locally, your submission runs in a Docker container; thus, there might be small differences to running it on AWS later. However, we are fine with such small installation troubleshooting as long as running the benchmarks then goes smoothly! :)