That being said I always try to write them in attribute style as it looks almost as clean as direct public access.
No you don't. You just need to read the published paper and attempt to reproduce what the paper reports. (A good scientific paper includes enough information to make the work it reports on reproducible.)
That's the ideal and idyllic world. In many areas there's a ton of hidden parameters and secret sauce that doesn't get reported. I found this to be especially true in robotics and machine vision and other computer science papers.
A freelance is one who gets paid by the word -- per piece or perhaps. -- Robert Benchley