Okay, let's operate on the assumption that blackbody radiation is the only viable and reliable (not going to consider the number points of failure involved in a heat barge idea- it's pretty to think about, but hardly practical) method of waste heat disposal. Mathematically, the amount of heat transferred via blackbody is proportional to the temperature of the emission surface and to the emissivity of the exposed surface. We've just about reached the technically feasible ceiling on the latter, but we might be able to squeeze out a little bit more emissivity with newer materials, which would allow us to shrink the required radiator space. The former, however, provides considerable room for improvement. Historically, spacecraft have used one waste heat removal system (albeit sometimes with redundancy). However, if we are to seriously consider using nuclear power as our primary source, we should consider designing a separate radiator system for the reactor and associated systems, one designed to function at much higher temperatures than would be acceptable for habitable modules, and keeping the structural interface between the reactor module and the habitation module at a minimum to reduce heat transfer. This also has the added bonus of reducing the amount of radiation shielding needed between the reactor and the habitation modules.
Assuming we can design a system that can operate reliably at twice the absolute temperature of the habitation modules, we can reduce the required heating surface (for the reactor, anyway) by a factor of 16, which assuming EyeR's calculations are correct, would mean about 125 m^2 per megaWatt, not including redundancy. While this would still necessitate a huge amount of surface area for our needs, it is still considerably more practical than previous concepts. We may reap further benefits from using multiple reactor modules, though I'd need to look at the practicality of this.