Scaling Pretraining Tasks in CodeT5 Narrows the Generalization Gap Across CWE Vulnerabilities

This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does scaling the number of pretraining tasks in CodeT5 influence its generalization gap between seen and unseen vulnerability types in the CWE dataset. 12 claims were extracted from source literature; 12 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.1/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: How does scaling the number of pretraining tasks in CodeT5 influence its generalization gap between seen and unseen vulnerability types in the CWE dataset?

Autonomous literature synthesis. Automated review score: 8.1/10. Full text and citation available at Assignee Research.