CodeFort: Robust Training for Code Generation Models

TL;DR

CodeFort is a framework that enhances the robustness of code generation models by employing diverse data augmentation and training strategies, significantly improving their resistance to code perturbations and errors.

Contribution

We introduce CodeFort, a comprehensive framework that generalizes code perturbations and integrates multiple robust training strategies to improve code generation model robustness.

Findings

Increased robust pass rates from 14.79 to 21.74.

Reduced robustness drop rate from 95.02% to 54.95%.

Supports high-throughput training with various strategies.

Abstract

Code generation models are not robust to small perturbations, which often lead to incorrect generations and significantly degrade the performance of these models. Although improving the robustness of code generation models is crucial to enhancing user experience in real-world applications, existing research efforts do not address this issue. To fill this gap, we propose CodeFort, a framework to improve the robustness of code generation models, generalizing a large variety of code perturbations to enrich the training data and enabling various robust training strategies, mixing data augmentation, batch augmentation, adversarial logits pairing, and contrastive learning, all carefully designed to support high-throughput training. Extensive evaluations show that we increase the average robust pass rates of baseline CodeGen models from 14.79 to 21.74. We notably decrease the robustness drop rate from 95.02% to 54.95% against code-syntax perturbations.