Advances in Code Translation and Optimization with Large Language Models

The field of software engineering is witnessing significant advancements with the integration of Large Language Models (LLMs). Recent developments indicate a shift towards leveraging LLMs for code translation, optimization, and repair. Researchers are exploring innovative methods to enhance the capabilities of LLMs in these areas, including data augmentation, rule-based analysis, and hybrid code editing. The focus is on improving the accuracy, efficiency, and effectiveness of LLMs in software engineering tasks. Noteworthy papers in this area include: SIADAFIX, which proposes an adaptive program repair method using slow and fast thinking, achieving state-of-the-art results. SemOpt, which introduces a novel framework for LLM-driven code optimization via rule-based analysis, demonstrating significant improvements over baseline methods. FidelityGPT, which enhances decompiled code accuracy and readability by detecting and correcting semantic distortions, showing substantial gains in accuracy and readability. PEACE, which presents a hybrid framework for project-level code efficiency optimization, achieving superior results over state-of-the-art baselines.

Sources

Automated Snippet-Alignment Data Augmentation for Code Translation

An Experimental Study of Real-Life LLM-Proposed Performance Improvements

SIADAFIX: issue description response for adaptive program repair

SemOpt: LLM-Driven Code Optimization via Rule-Based Analysis

When Many-Shot Prompting Fails: An Empirical Study of LLM Code Translation

PEACE: Towards Efficient Project-Level Efficiency Optimization via Hybrid Code Editing

FidelityGPT: Correcting Decompilation Distortions with Retrieval Augmented Generation

BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills

Large Language Models for Fault Localization: An Empirical Study

Built with on top of