> or in computer science / applied math/stats / etc., with different codebases, with different model variants, on different datasets) and the overall conclusions hold
A lot of open sourced CS research is not reproducible.
"the code still runs and gives the same output" is not the same as reproducibility.