Are LLMs actually capable of identifying relevant constructs in academic papers?

I compare output generated by Large Language Models analyzing academic writing to identify predefined constructs against humans analyzing these same pieces of writing.

to follow