I usually don't see path loss models for IRS consider polarization. But specific to mmWave application, the polarization of the antenna has some effect. Is there any research that has explored this?
If IRS means intelligent reflecting surface, then work done at cm (microwave) wavelengths will be relevant at mm wavelengths. I think there will be a lot of material about polarization dependence of reflections from all sorts of surfaces at microwave frequencies. Just scale the thickness and other sizes, and usually there is more loss at mm. Simpler structures tend to be used at mm, because of the difficulty of building smaller structures, so that similar work at microwave tends to be at least a few years old.