In this paper, Dense Passage Retrieval (DPR) is cited as historical background for modern dense retrieval and RAG, not as the literal implementation of the paper's flat baseline. The paper's actual flat dense baseline uses the shared dense seed-pool setup and strict-parity evaluation contract: the encoder, candidate pool, cutoff $k$, split, and matching rule are fixed so graph-aware methods can be compared against a non-graph retrieval condition. This node therefore teaches DPR's role as background context while leaving the actual flat-baseline details to the separate flat dense retrieval baseline node.