Graph-constrained path selection on contextual keyword centroids decouples path enumeration from teacher verbalization to scale multi-hop QA data from plain text, enabling 4.4x corpus expansion and raising closed-book Token F1 from 21.66% to 38.58% on CUAD legal contracts.
find candidate evidence paths that the teacher comfortably verbalizes
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Scaling Multi-Hop Training Data via Graph-Constrained Path Selection
Graph-constrained path selection on contextual keyword centroids decouples path enumeration from teacher verbalization to scale multi-hop QA data from plain text, enabling 4.4x corpus expansion and raising closed-book Token F1 from 21.66% to 38.58% on CUAD legal contracts.