> LLMs are famously inefficient, expensive and unreliable.
Maybe, but here I am suggesting an LLM specifically trained to do only machine code to high-level code reverse engineering.
I suspect this can be made both efficient and very reliable.
Especially given the fact that there is very often a path the verify if what the LLM produces is a valid answer (by testing if the high level code produced by the LLM, when compiled, behaves like the original machine code).
Maybe, but here I am suggesting an LLM specifically trained to do only machine code to high-level code reverse engineering.
I suspect this can be made both efficient and very reliable.
Especially given the fact that there is very often a path the verify if what the LLM produces is a valid answer (by testing if the high level code produced by the LLM, when compiled, behaves like the original machine code).