Yes there are some really interesting projects, also in the ML replicability space.
One really nice approach is the DAT project [1]. The protocol [2] looks pretty sensible and useful. Unfortunately, the tooling has been in such a state of permanent flux (i.e. perpetual deprecation) that I've never bothered to invest much time.
[1] https://datproject.org/
[1] https://datproject.org/