I don't think the point is compression, SDFs even when represented with a neural net have a much larger memory footprint and are significantly more expensive to render.
The main benefit of a SDF are that you don't have topological issues to worry about, which makes them great for reconstruction problems where a surface is evolving over time.
The main benefit of a SDF are that you don't have topological issues to worry about, which makes them great for reconstruction problems where a surface is evolving over time.