I want to use "eigs" on a Hermitian matrix. If I have a symmetric matrix, I can tell "eigs" that it is symmetric (and real) and it will take advantage of this. But if it is Hermitian, it seems "eigs" does not exploit this information (and it definitely should be able to, since we know the eigenvalues are real, for example). Is this due to a limitation in ARPACK (which I think is the underlying solver), or a limitation with Matlab's wrapper? Any work-arounds?