GraphProof-QA: Teaching a Small Model to Prove Its Answers

A 1.5B model that writes an executable graph query instead of answering from memory: 93.3% vs 34.2%, and 81.4% vs 5.8% on unseen names.