Hi @CaioSaporito ,
I think the bending moment signal is correct, as with the negative gradient, you would expect the "colder" temperature to be on the top of the panel and the "warm" temperature to be on the bottom of the panel.
The top of the panel (colder temp) is trying to contract inward, but because the edges are pinned, it can't. So tension forces are generated on the top. On the other hand, the bottom of you panel (warm side) is trying to expand but it can't due to the pin support, thus a compression force is generated at the bottom.
However, your supports provided are only "pinned", so theoretically speaking you should be getting "ZERO" bending net in the slab as the negative side of the grading would cancel out with the positive side given your slab is FREE to ROTATE at the edges.
I'm only guessing, but I would think the small bending you getting is due to the shape functions error in Robot or maybe you getting some minor bending restrains at the corners due to the slab being supported on all sides.