I think I have found a decent solution to this. It uses Yahoo Pipes on the RSS feed from arxiv.org. Have a look at this screenshot from the interface:

It is as simple as telling Pipes to get the feed, filter it according to your needs and providing the output which you can in turn prenumerate with your favourite RSS or other news aggregator. I have created three such pipes for myself, one for a list of author names, one for keywords in the title or abstract and one for certain favourite astronomical objects that I work with.

This has reduced the number of papers that I have to check to around ten per day and the chances of them being relevant are usually quite good. Of course the results depend completely on your filter criteria and this whole thing might not be for you if you are afraid to miss something.