I am aware that classifying all $SL_2(\mathbb{Z})$ representations is more or less completely intractable, but I was wondering what is known about the following simpler question: How do I recognize when a representation $SL_2(\mathbb{Z}) \to GL_N(\mathbb{C})$ is the restriction of a representation of $SL_2(\mathbb{C})$?

 Ideally I'd like the criteria to be some finite collection of properties of how finitely many specific elements of $SL_2(\mathbb{Z})$ act. Here's an example of what I have in mind: If we ask the analogous question for $SL_n(\mathbb{Z})$ with $n\ge3$, then a representation $SL_n(\mathbb{Z}) \to GL_N(\mathbb{C})$ comes from a representation of $SL_n(\mathbb{C})$ if and only if an elementary matrix acts unipotently.

A few things I have found during a brief search of the literature:

 - There is a low dimensional classification of representations of Tuba and Wenzl in "Representations of the Braid Group $B_3$ and of $SL(2,\mathbb{Z})$". It seems that the unipotency condition is not sufficient when $N = 5$ (obviously it is still necessary), but there are only finitely many other representations with this property.
 - If there is a representation of $SL_2(\mathbb{Z})$ with a compatible action of the Borel $B_2(\mathbb{C})$ then a result of Demazure in  "Groupes reductifs de rang semi-simple" says they extend to $SL_2(\mathbb{C})$. This is nice, but I'd still want criteria for when such an action of the Borel can be constructed.

Is this question addressed somewhere in the literature?