As a non-moderator, I think it's sufficient to write a comment asking for clarification & register a close vote at the same time. If a comment already asks for clarification, and you agree with it, then you only have to up-vote the existing comment.
The reason I say "non-moderator" is that it takes 4 other non-moderator votes to close the question; perhaps it's unclear to you but clear enough to other reviewers.
I say that you should vote to close at the same time because it helps precisely no one to have an open & unclear question. If OP cares enough, they will revise it. If the question is revised before it's closed, you can retract your close vote. If the question is revised after it's closed, you can vote to re-open. I try to check on the status of the questions I've recently voted to close as unclear (sometimes the edits help a lot but the re-open vote gets stuck at 4 votes) but this is admittedly imperfect.
The trick is to write a comment that is precise enough that OP knows what kind of clarification to add. Sometimes I get this wrong and I ask for clarification in a way that invites OP to add a bunch of irrelevant context.
Sometimes I also delve into old, unanswered questions that are not closed and which bear a particular tag. For example, I might query
[neural-networks] answers:0 closed:no and then sort by date, starting with the oldest questions. The purpose of this exercise is to remediate the question in some way -- upvote & answer it, find a duplicate, or close it. Less commonly, if it's on-topic and not a duplicate and I don't know the answer, then I just up-vote it and move on; not everyone can answer every question!
To do this efficiently, I've outlined a mental procedure that I walk through for each question. These are not in any particular order.
- Is this an answerable question? (For example, questions that ask "is a neural network or logistic regression better?" or "what threshold should I choose for my classifier?" is not an answerable question.)
- Can a person who is familiar with this topic write a clear, concise answer to it? Or is critical information missing?
- Does this question ask about statistics or machine learning as discussed in the help center?
- Does answering this question rely on writing code or debugging code?
- If I search stats.SE for the key words, do I find a duplicate question?