Graph-based Evidence Verification and Explainable RAG Framework for Space Missions

This paper proposes a fresh approach to checking facts uses graphs, built into a clear step-by-step process meant for trustworthy help in space missions. Instead of just pulling data, it finds official manuals, checks details through linked logic, then answers with reasons shown. What sets it apart is how it says no when unsure - no guesses made. Running without internet access, it relies on a freely available tuned language model that follows directions well. Tests show fewer mistakes, less made-up output, better visibility into decisions than older methods. A fresh approach called Retrieval-Augmented Generation pipelines now helps LLMs tap into outside information. Still, current versions tend to invent details, offer no clear reasoning trail, nor allow checks on accuracy. Because of these gaps - like unverified outputs and opaque logic - they struggle where trust matters most. Think spacecraft operations, where mistakes aren't an option. This paper proposes a method to work offline critical missions.