This depends on how you look at it.
In the standard cosmology we define a preferred frame wherein an observer will see the matter (that's in the most general sense of "not the gravitational field", so it includes atoms and their components, photons, and various types of dark matter (e.g. neutrinos, which are "hot" dark matter, since they move relativistically and do not experience electromagnetism)) content of the universe as homogeneous and isotropic. This is physically reasonable since along every unobscured line of sight we see a lot of galaxies of various shapes, "tilts", sizes, surface brightnesses, and spectral lines. Observations also lead us to conclude that there is a relationship between redshifting of the spectral lines of common types of galaxies (and common radiative occurrences within them, like type A supernovas), and the change of the other observables (angular size on the sky, luminosity, etc.) that correlate with greater distance. This in turn led to the discovery of the Hubble "constant", and provoked ever deeper field telescopic studies to prove its value.
So if we assume that along every line of sight, including obscured ones, we have much the same view of many many galaxies at a variety of distances, we can make a variation on the Friedmann equation that parameterize several things that would lead to the observables of galaxies when we model their known (and unknown) components as a set of perfect fluids.
We can take the Hubble "constant" and put it into a Robertson-Walker vacuum spacetime. RW spacetimes can be grokked by dimensional reduction. Consider a cylinder that we slice (foliate) along its axis into a set of infinitesimally thin circles stacked on top of each other. We describe the radius of each circle with a function r(h) where h is the height of the circle from the base of the cylinder. Where r(h) is constant, we have a cylinder, but if r(h) increases with h, then we have something like a cone balancing on its apex; r(h) can describe a wide variety of shapes. For an 3+1 RW spacetime that is similar to our universe, we foliate on the timelike axis and define a function a(t) where t_0 == now with the spacelike coordinates set on a chosen observer (us here on Earth, for example). "t" counts upwards as we go into the past from t_0, and a(t) goes to zero as t increases. "t" is the lookback time and a(t) is the scale factor. The chosen observer is the special observer mentioned above, who sees the matter of the universe as isotropic and homogeneous at the largest scales. The most useful coordinates on such a spacetime are comoving, that is each gravitationally bound galaxy cluster stays at the same coordinates at every time t.
If we mix together the Friedmann equations, the Lemaître idea of spacetime having a zero radius at some large lookback time, and the RW spacetime that can model that, we get the FLRW model of the standard cosmology. We take the RW case where there is no extrinsic curvature, that is, when we foliate on the timelike axis each spacelike hypersurface is spatially flat; that is similar to saying that when we slice up our dimensionally reduced solid along its height, we get a set of circles of the same radius (i.e. a cylinder rather than a cone). We absorb the expansion parameter into a(t) as another of the fluids.
We then consider two types of fluid: those that dilute away as t -> t_0 -> future and those that do not dilute away. The former is "matter", including dark matter; the latter is "dark energy".
When we consider them as components of an action, diluting-away fluids are attractive and non-diluting fluids are repuslive. When we consider them in terms of the matter tensor T in General Relativity, the diluting-away fluids have positive pressure and the non-diluting fluids have negative pressure.
It's important to return to the point that this model has a preferred frame, and that translating the non-diluting fluid into "the same for all observers in all frames of reference" physics leads one to assume that dark energy is just a feature of the Lorentz-invariant vacuum. So dark energy arises in the cosmological model but corresponds to the ground state of the empty-of-matter spacetime in frames of reference other than the preferred one picked out by the cosmological model.
That is, the statement that "dark energy drives the expansion (via negative pressure or repulsion)" is frame-dependent, and thus observer-dependent, and with a change of frames of reference (and even a change of coordinates on the preferred frame), the statement becomes untrue. What is true in all frames is that there is an intrinsic property of space in an expanding spacetime that has a constant energy-density no matter how large a volume of space is considered.
The exact equation of state of dark energy is an area of active research and also tests to make sure that the assumptions that inevitably lead to it (isotropy at huge scales, homogeneity, spatial flatness, redshift-distance relations and other things implying expansion) are not blown up by evidence from ever finer observations.
So it's not so much a 'label' as a phenomenon whose microscopic details have yet to be discovered.
There are lots of those in physics, and we've had a good century of probing the microscopic details of phenomena discovered at the end of the 19th century and since, so this shouldn't really be causing anyone sleepless nights.