[roel_v]

"There are some open source websites that are trusted (e.g. https://arxiv.org"

Lol. Citing arxiv is even worse than citing wikipedia, and some reviewers will reject arxiv citations all together. It's full of junk 'science' by crackpots. Not just 'hey look I ran this regression on a public dataset' bad, but all-out 'I was abducted by aliens, and I made up an equation to show that they took me to Pluto' bad. Maybe it differs by field, I don't know - but suggesting arxiv is 'trusted' the same way Nature is 'trusted' is inane.

[lorenzhs]

Yeah, nobody would consider an arXiv-only article by an unknown author. There are plenty of P=NP "proofs" published on the arXiv. But that isn't what it is for. It's a preprint server where you have to do your own quality control. A lot of work in the theoretical CS community is published in conferences and on the arXiv, often in extended form (additional proofs, plots, etc that didn't fit into the conference publication's page limit)

From https://arxiv.org/help/general: "Disclaimer: Papers will be entered in the listings in order of receipt on an impartial basis and appearance of a paper is not intended in any way to convey tacit approval of its assumptions, methods, or conclusions by any agent (electronic, mechanical, or other)."

[tomw2005]

I was not meaning to equate ArXiv to Nature. I can see how I implied that though.

As I've said in another comment it was used much more frequently during my undergrad years than at postgrad. Within Maths and Physics it is generally trustworthy as they are pre-prints of papers to established journals.

Would I cite an arxiv reference in a paper for submission or thesis? 99% of the time, no. Useful for citing on the web though where your users may not have access to the final journal article behind a paywall.