I've been reading papers from ICSE'07 (now that I'm a prof, I figure it's part of the job), and have mostly been underwhelmed. As a grad student I work with commented upon returning from the conference, there doesn't seem to be much connection between academic research in software engineering, and what good software developers actually do, use, or care about.
That said, I was very excited by Zimmermann, Premraj, and Zeller's paper "Predicting Defects for Eclipse". They have mapped defects from Eclipse's bug database to source code locations for the 2.0, 2.1, and 3.0 releases. Even better, they've put their raw data on the web at http://www.st.cs.uni-sb.de/softevo/bug-data/eclipse/ so that other people can use it in their own studies and experiments. I really wish we were able to do the same with student code repositories, so that people who are mining CVS and SVN to find patterns in development could test their algorithms, but privacy rules forbid it.