Ok then. you have just said that gravity can lead to large filaments in space. Can you explain how this is possible using an exclusively attractive field?
On one hand, what's to explain? These research groups just put 1,000,000 particles in a simulated box, made them all attract one another with F= ma = F = Gmm/r^2, and the result is filaments. This is not a fancy new result; it's the only way that these equations work, and it's been around for decades. So, yes, filaments are possible with exclusively attractive gravity.
On the other hand: yes, I can explain. The initial matter distribution was fairly, but not perfectly, smooth. Regions with slightly higher densities then attract everything else. Particles start falling towards the overdense regions. Any individual particle, though, will start moving towards the nearest overdensity it can see. Since there are overdensities on all scales, this behaves a lot like dividing space into Voronoi volumes, and everything is attracted towards the nearest boundary. At the end of this process, you've got a "sponge" in which everything has congregated on the walls. Where two walls intersect, the density is higher than the density in mid-wall; attraction of the mid-wall particles to the wall-intersection particles draws everything down into the filaments.
This happens yet again at the intersection of two filaments. The particles in the filaments are being attracted to one another but also towards the big mass at the nearest intersection. Given enough time, the big point masses at the ends of each filament will attract all of the filament particles, and then we'd have the Universe you seem to expect---just a bunch of point masses.
This is an approximate description, because all of these processes are going on at the same time rather than one after another. Moreover, the Universe is expanding faster than the superclusters can draw material out of the filaments, so the present structure is probably "frozen in".
Don't believe my handwaving description? That's fine. Believe the numerical solutions instead.