CodeT5 generally outperforms CodeBERT on complex vulnerability datasets, but both models' fixes degrade sharply on out-of-distribution data.
In: 8th USENIX Conference on Operating Systems Design and Implementation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Multi-Dataset Evaluation of Models for Automated Vulnerability Repair
CodeT5 generally outperforms CodeBERT on complex vulnerability datasets, but both models' fixes degrade sharply on out-of-distribution data.