I would imagine it depends on the particular paper, the more experimental ones in CS, the more theory ones in math. Do note though that most math and ML practi

I would imagine it depends on the particular paper, the more experimental ones in CS, the more theory ones in math. Do note though that most math and ML practitioners use arxiv over sci-hub.