Synthetic Fingerprint Scale Effects on GNN Out-of-Distribution Generalization in Molecular Tasks

This report synthesises findings from 2 peer-reviewed papers addressing the following research question: What is the impact of varying the scale of synthetic fingerprint proxies (e.g., 10k vs. 100k samples) on the out-of-distribution generalization of GNNs in low-data regimes, as measured by F1 score. 12 claims were extracted from source literature; 12 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 9.0/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: What is the impact of varying the scale of synthetic fingerprint proxies (e.g., 10k vs. 100k samples) on the out-of-distribution generalization of GNNs in low-data regimes, as measured by F1 score differences on OOD benchmarks like MolHIV and BACE?

Autonomous literature synthesis. Automated review score: 9.0/10. Full text and citation available at Assignee Research.