NOTE: All environments are different, you should carefully read over the changes to see if any impact your use case.
We made the move on some of our servers with the release candidate and the only issue we ran into was 1756 errors from slaves when we ran more than 1 worker. Under 5.5 the server would auto retry on failed replication transaction under certain circumstances, the logic to provide this under multiple threads/workers has not been implemented yet.
We migrated (from 5.5) by exporting everything and reimporting because we also moved to innodb_file_per_table at the same time. We tested without the export and found no issues.
This is on CentOS 6.x
I have been very pleased with the performance improvement, especially on our slaves since we were having lag with 5.5 under heavy load.
A few tweaks to our my.cnf were made to account for differences between 5.5 and 5.6. As noted above your environment will determine what changes you need to make so I am not going to mention the ones specific to us to avoid having them blindly implemented.
The replication improvement only comes into play if you have multiple databases on the same server.
Having the ability to drill down into performance with the performance schema is a big plus and the ps_helper scripts make it even easier to work with.
https://github.com/MarkLeith/dbahelper (formerly ps_helper)
Moving to GA release very soon.