Let's start from the beginning. If you look at a single atom, you find distinctive energy states for an electron, but in solid state physics these distinctive energy states blur into continua of possible state, called bands.
Usually you will find plots of such band structures in the momentum space and here the mass of the electrons come into play. Juggling a bit with formulas one can see, that the curvature (second derivation) of the band structure is the (effective) mass of that particle.
So, is this a "real" mass? Well, who cares, but certainly the particles behave like they have this mass and that's also why polaritons are so interesting.
If you bombard your electrons with photons, so that they absorb/emit with an high enough rate (strong coupling), they stop behaving like electrons and instead behave like a combination of photons and electrons. We call this quasiparticle a polariton.
Now, to tie that all together, the photon has m=0 and since the polariton inherits some properties from it, it also has a drastically reduced effective mass.
The guys from this paper seem to have found a configuration, where the effective mass is not only small, but negative, which is surprising.