QASC Rebuilt as Directed Science-Fact Graph (16,444 Nodes, 25,590 Edges) Used Only for Validation Retrieval
In this paper, QASC is rebuilt as a directed science-fact graph with nodes and edges. Graph descriptors computed from the released graph and fact texts are: mean depth , fraction of nodes with depth equal to , mean degree , and mean fact-text length tokens. Because the official test split withholds evidence, the paper reports only validation retrieval on the 926-question validation split. The supplement Datasets section documents the node-extraction rule, edge-construction criterion, and directionality assignment that reproduce these counts exactly. QASC functions as a distinct-domain (non-prerequisite) fact-graph control rather than as a headline prerequisite benchmark.
0
1
Tags
Science
Auditable Strict-Parity Evaluation of Prerequisite-Graph Retrieval for RAG under Leakage Controls
Related
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
QASC Rebuilt as Directed Science-Fact Graph (16,444 Nodes, 25,590 Edges) Used Only for Validation Retrieval