TY  - RPRT
TI  - Graph-Native Reinforcement Learning Enables Traceable Scientific Hypothesis Generation through Conceptual Recombination
AU  - Subhadeep Pal
AU  - Shashwat Sourav
AU  - Tirthankar Ghosal
AU  - Markus J. Buehler
PY  - 2026
UR  - https://arxiv.org/abs/2607.00924
ID  - 2607.00924
ER  -