AST-T5: Structure-Aware Pretraining for Code Generation and Understanding
This addresses the need for better code generation and understanding in AI, offering a novel approach that is incremental but provides specific gains in code-related tasks.
The paper tackles the problem of large language models neglecting code structure by introducing AST-T5, a pretraining method that leverages Abstract Syntax Trees, resulting in improved performance such as surpassing CodeT5 by 2 points in exact match for Bugs2Fix and 3 points for Java-C# Transpilation.
Large language models (LLMs) have made significant advancements in code-related tasks, yet many LLMs treat code as simple sequences, neglecting its structured nature. We introduce AST-T5, a novel pretraining paradigm that leverages the Abstract Syntax Tree (AST) for enhanced code generation, transpilation, and understanding. Using dynamic programming, our AST-Aware Segmentation retains code structure, while our AST-Aware Span Corruption objective equips the model to reconstruct various code structures. Unlike other models, AST-T5 avoids intricate program analyses or architectural changes, so it integrates seamlessly with any encoder-decoder Transformer. Evaluations show that AST-T5 consistently outperforms similar-sized LMs across various code-related tasks. Structure-awareness makes AST-T5 particularly powerful in code-to-code tasks, surpassing CodeT5 by 2 points in exact match score for the Bugs2Fix task and by 3 points in exact match score for Java-C# Transpilation in CodeXGLUE. Our code and model are publicly available at https://github.com/gonglinyuan/ast_t5.