For conferences
Check that submitted results came from the submitted code
Reviewers can't rerun most experiments. They take the numbers on trust.
What it helps with
No re-running
Check each number against the code that produced it.
Double-blind ready
Reports carry no author names.
Every run counted
Failed and discarded runs show up, not just the best one.
Paper vs results
Claims in the paper are checked against the sealed numbers.
Artifact evaluation
Confirm the artifact is the one behind the results.
How it works
Authors
Run experiments with kveritas, seal, and attach the report and code.
Reviewers
Upload the report at kveritas.org/verify. No account needed.
Reviewers stop asking whether the code really ran, and spend their time on whether the work is right.
Frequently asked questions
Does it prove the results are correct?+
No. It proves they came from the sealed code, unchanged. Judging the work stays with people.
Does it work with double-blind review?+
Yes. Reports are anonymous by default.
Do authors have to share their data?+
No. The data is never read or uploaded.
What does it cost authors?+
One command prefix, and under 1% extra runtime in our tests.