What should I read to understand the theory behind multi stage ML?
The meta-family of language variants (MetaML, MetaOCaml, MetaScheme, and MetaHaskell) each have some associated literature. I’d start with MetaML and MetaOCaml on Google Scholar.
Oleg has some fun stuff too: http://okmij.org/ftp/meta-programming/
You can skim that page and work backwards from keywords in there.