Yet. You’re adding one other thing that authors need to keep in mind when developing the product, fixing bugs, and adding features. The fact that the input must be trusted is not an intrinsic part of the business logic, it’s an additional caveat that humans need to remember.
Well think about this from a product perspective. A natural extension of this is to be able to simulate tax code that hasn’t been implemented yet. “Bring your own facts” is practically begging to be a feature here.
That repetition of variable and name is not the most terse, though. At least with XML, the repetition in the end tag is handled for you by pretty much every XML-aware text editor.
The code that processes the graph will also be simpler as you don't have to parse the XML graph and turn it into something that can be executed.
And if you still need XML, you can generate it easily.