- the WZW model in 2 dimensions and
- the Chern-Simons theory in 3 dimensions.
A success story of this proposal is that one to give a much more geometric and dynamical description of Verlinde's formula, Cardy's condition, and other things. The details of this approach are different from the "N=2" string strategy - the latter offers a higher probability to explain the exceptional duality etc. - but still, some of the results look nontrivial. There are many open questions, too. For example, the original of the conformal bc ghosts is not really understood.
Anyone is familiar with this direction of thinking?