GROMACS version:2019.1
GROMACS modification: No
Here post your question
Theory question: I am conducting simulations in vacuum and I see unreasonable structural fluctuations with Nose-Hoover thermostat but not with Langevin. Why would those additional frictional forces added by Langevin dynamics are at an advantage when each molecule is kept in a very large box .i.e. when molecules are not interacting with each other?
With stochastic dynamics each degree of freedom is thermalized individually. This results in near perfect equipartitioning of kinetic enery. With Nose-Hoover it can take an extremely long time for kinetic energy to redistribute, especially for translational and rotational degrees of freedom of molecules without solvent (which you seem to have?).