Induced HotpotQA Slice as Auxiliary Stress Test in Supplement
The paper's supplementary materials use a denser induced slice of HotpotQA as an auxiliary stress test only — not as a headline benchmark. The slice is constructed from HotpotQA so that its induced relation structure is denser than the curated prerequisite DAGs, which probes whether graph-aware retrieval behavior changes as graph density increases. Because the slice is supplementary, it cannot support headline retrieval claims; it functions as an external-validity boundary check.
0
1
Tags
Science
Auditable Strict-Parity Evaluation of Prerequisite-Graph Retrieval for RAG under Leakage Controls
Related
Induced HotpotQA Slice as Auxiliary Stress Test in Supplement
HotpotQA FullWiki-1k Boundary Probe: Flat 93.4, Hierarchical 92.9, Adaptive 94.0 R@10
HotpotQA External-Validity Probe: Adaptive Depth Does Not Transfer to a Denser Non-Prerequisite Graph (FullWiki-1k: Flat 93.4 / Hier 92.9 / Adaptive 94.0 R@10)
LectureBank-Full Configuration Used in Hierarchical Prerequisite RAG (208 Concepts, 899 Edges, 1,421 QA)
MOOC-CS Configuration Used in Hierarchical Prerequisite RAG (225 Concepts, 516 Edges, 1,016 QA)
Canonical Prerequisite Splits Are Heavily Templated: 92/80 LectureBank-Full and 68/60 MOOC-CS Train-Test Overlaps
QASC Rebuilt as Directed Science-Fact Graph (16,444 Nodes, 25,590 Edges) Used Only for Validation Retrieval
Induced HotpotQA Slice as Auxiliary Stress Test in Supplement