First measured result on whether epistemic metadata changes how AI consumers answer derivation questions over an MBSE model. 3 LLMs × 120 judged answers over the public Apollo 11 SysML v2 model. Anonymised for double-blind review.
reproducibility ai-safety systems-engineering apollo-11 mbse hallucination llm-evaluation confabulation systems-engineering-ai sysml-v2 epistemic-adequacy
-
Updated
Jun 28, 2026 - Python