For any particular subfield of CS (Machine learning, Computer Vision, HCI, etc.), find some top universities that offer graduate seminars, find the website which will usually list a reading list.
Completely agree with this approach. Conferences, at least for me, are not that helpful. Grad level advanced courses are usually the best resource because other people have "filtered" interesting content and organized it into some common theme e.g, I recently took Kai Li's course on "Big Data":
For example, from UW:
Machine Learning: http://courses.cs.washington.edu/courses/cse590m2/09au/ HCI: http://courses.cs.washington.edu/courses/cse590h/13au/
I've found graduate reading seminars usually try to present a mix of seminal, eclectic and recent that gives you a diverse overview of the field.