I'm curious how does credit work in these situations? Is it first to upload/publish somewhere (arvix) or is it "here are the logs when we found it"? If it's the

I'm curious how does credit work in these situations? Is it first to upload/publish somewhere (arvix) or is it "here are the logs when we found it"? If it's the former, I'd build a pipeline that autogenerates a quick paper and uploads immidiately when the number is spit out. If it's the latter, how do we make sure the logs are not tampered with? Either way it's fun for me that one of the teams used the Noctua 2 cluster, I was in the server room a couple of month ago. Looked very underwhelming but it's quite interesting because I've never seen an FPGA cluster before. They are also doind some interesting ML things there (optimizing for power consumption). Edit: nevermind the first part, I read the van Hirtum paper and they registered the number in Github. Can't find any other time mentioned in the Jäkel paper.