I'm a little skeptical of the results as well. A tonne of CO2 releases 2489 kW*hr of free energy when formed from C and O2.

However, reading the original paper, it seems like the end product is "highly oxidized carbon", "akin to that of slightly crystalline graphene oxide". So they're not claiming to undo the carbon oxidization, more to vitrify the carbon oxide in a solid phase. This, presumably, takes significantly less energy than reducing the carbon.