The change that leads to this behavior is that we now HTML-encode texts. We do this as SVG is not our only rendering target - we support Canvas and PDF as well. These will output the encoded text as they don't have a notion for HTML entities. If we don't encode the text we'll get different results, e.g. during PDF export.
What we suggest instead is to use visual templates for the labels. In combination with the Layout primitive we can produce the same results and much more.
I've prepared a snippet to demonstrate your particular case. Does this make sense?
Join us on our journey to create the world's most complete HTML 5 UI Framework - download Kendo UI