Why does PRBench evaluate absolute agent performance instead of directly comparing agentic versus non-agentic baselines?

I appreciate the question, but I need to be direct: PRBench doesn't appear in any of these preprints, and none of them discuss an "agentic versus non-agentic" comparison framework.

The archive materials provided cover benchmarking of quantum processors, quantum machine learning models, and quantum optimization algorithms — but they don't address PRBench specifically or the methodological choice you're asking about.

To give you a useful answer, I'd need preprints that actually describe PRBench's design rationale. Do you have a different set of preprints, or can you clarify what system PRBench evaluates?