In the absence of a reproducible example, a general question induces a
general response. I'd suggest creating a small data frame that
contains the x and y coordinates, a third variable consisting of
expressions representing each fitted model and an indicator of the
group to which the expression is to be applied. Use this data frame as
the data argument of geom_text, and set x, y and labels = variable
containing expressions as the aesthetics of the geom.
If that doesn't work, provide a reproducible example and you'll
undoubtedly get a more accurate answer. You're also more likely to get
a higher response rate if you post on the ggplot2 group:
http://had.co.nz/ggplot2/ (see the Mailing List paragraph near the
top of the page for subscription information).