As a reader of papers on Arxiv, I get a lot of value from having author names and institutions on them, and having them all be by "a group of machine learners"

As a reader of papers on Arxiv, I get a lot of value from having author names and institutions on them, and having them all be by "a group of machine learners" would be a pain. I want to be able to find high quality work that's relevant to my interests, and looking for papers by authors whose work I've liked in the past is a very effective way to do that. Obviously if I'm a conference reviewer, it'd be terrible if I used this same heuristic, and so I'm sympathetic to the fact that author anonymity does have real benefits. But we shouldn't just ignore the fact that author identification also has real benefits.