Advances in Code Generation and Translation

The field of code generation and translation is moving towards increased use of large language models (LLMs) to improve accuracy and efficiency. Researchers are exploring new approaches to ensure the correctness of generated programs, such as constrained decoding algorithms and multi-agent frameworks. These innovations have the potential to significantly impact areas like robotics and legacy system modernization. Notable papers include: Correctness-Guaranteed Code Generation via Constrained Decoding, which presents a novel decoding algorithm for generating semantically correct programs. RepoTransAgent: Multi-Agent LLM Framework for Repository-Aware Code Translation, which proposes a multi-agent framework for repository-aware code translation and achieves significant improvements over state-of-the-art baselines.

Sources

Correctness-Guaranteed Code Generation via Constrained Decoding

RepoTransAgent: Multi-Agent LLM Framework for Repository-Aware Code Translation

LaTeXTrans: Structured LaTeX Translation with Multi-Agent Coordination

GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging

An LLM-powered Natural-to-Robotic Language Translation Framework with Correctness Guarantees

MultiPL-MoE: Multi-Programming-Lingual Extension of Large Language Models through Hybrid Mixture-of-Experts

Leveraging LLMs for Automated Translation of Legacy Code: A Case Study on PL/SQL to Java Transformation

Built with on top of