In trying to construct a special relativistic model of quantum mechanics, one has to require invariance of the probabilities under Poincare' transformations.
Then if we let the spinorial wave function to be $\Psi_{k}(x) \in L^{2}(\mathbb{R}^{4}) \otimes \mathbb{C}^{2s+1}$ it seems we should have the following equation:
\begin{equation} \Psi_{b}(\Lambda_{\nu}^{\mu} x^{\nu}) = M_{ab}(\Lambda)e^{\frac{i}{2}\hat{L}_{\mu\nu} \omega^{\mu\nu}}e^{i\hat{p}_{\mu}a^{\mu}}\Psi_{a}(x^{\mu}) \end{equation}
Where $\Lambda$ is a Poincare' transformation. Now we can see that the spinor index has to transform non-unitarily, because the $M_{ab}(\Lambda)$ are finite-dimensional representations of a compact group.
However, to require invariance of the probabilities we must have $M_{ab}(\Lambda)M^{\dagger}_{bc}(\Lambda) = \delta_{ac}$ by taking the square of both sides, but this cannot be as the representation is finite. How is this problem overcomed in QFT?