ReAlignFit uses chemical induced fit bias and subgraph information bottleneck to dynamically align molecular substructure representations and improve stability on rule-shifted and scaffold-shifted data.
DGCL: Dual-Graph Neural Networks Contrastive Learning for Molecular Property Prediction,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Representational Alignment with Chemical Induced Fit for Molecular Relational Learning
ReAlignFit uses chemical induced fit bias and subgraph information bottleneck to dynamically align molecular substructure representations and improve stability on rule-shifted and scaffold-shifted data.