As someone who enjoyed X{ML,SLT, *} and RDF, I'm nonplussed with the idea that to get a JSONSchema conforming response from an LLM I need to resubmit the failing response with an additional “please for real this time” prompt.
As someone who enjoyed X{ML,SLT, *} and RDF, I'm nonplussed with the idea that to get a JSONSchema conforming response from an LLM I need to resubmit the failing response with an additional “please for real this time” prompt.