Unfortunately we don't see most of them. Pollsters working for candidates produce accurate polls or else they would not be working very long. But campaigns don't release them. If they are ahead they don't want their supporters to get too confident. If they are behind they don't want their supporters to peel off and they suffer an even worse defeat.

the difference between them. As we do not know more than they do who will vote, reacting at each individual poll is meaningless.

However, composite polls like Silver, Wang and others do are more useful. They have different methodologies, but right now, they all consider it is more likely that the GOP will win the Senate (but for most of them, it is more likely, but not a certainty).

But it depends on the details of the poll.

A whole lot of the polls we see are "likely voter" polls. Which means the results are directly affected by the pollster's model of who is likely to vote. If you think rural whites will turn out but urban minorities will stay home, you're gonna get a poll that shows a Republican win. If your prediction of "likely voters" was right, your poll will look accurate. If your prediction of "likely voters" was wrong, your poll will look wrong.

That's what the "Unskewing" Republicans did in 2012, and why they turned out to be so wrong - they decided Democratic constituencies would stay home.

Part of Nate Silver's shtick is to keep releasing polls, so that he can keep tweaking his "likely voter" model (He's also a decent statistician, which helps)