RKE-Coder: A LLMs-Based Code Generation Framework with Algorithmic and Code Knowledge Integration

Yajiao Deng

doi:10.1007/978-981-95-3343-5_41

RKE-Coder: A LLMs-Based Code Generation Framework with Algorithmic and Code Knowledge Integration

2026-01-01

0

PRE

AI

DOI：10.1007/978-981-95-3343-5_41

Original

Save

Abstract

En 中文

Code generation tools are crucial for improving the efficiency and quality of computer programming. However, existing tools lack a deep understanding of algorithmic knowledge, leading to low-quality code logic in code generation problems. In this paper, we propose a novel framework for code generation, named RKE-Coder, designed to leverage domain knowledge to generate high-quality code embedded with algorithmic insights. The framework aims to enhance the quality and accuracy of code generation for computer programming problems with two types of representative knowledge: algorithmic knowledge and code knowledge. Experimental results show that the RKE-Coder successfully incorporates algorithmic knowledge into the generated code, improving the accuracy and quality of code produced by the LLMs for computer programming tasks. This research makes a significant contribution to competitive programming code generation. It provides a more powerful tool for programmers to address the ever-evolving challenges in computer programming problems.

Keywords:

Code Generation

Algorithmic and Code Knowledge

LLMs

RKE-Coder: A LLMs-Based Code Generation Framework with Algorithmic and Code Knowledge Integration

Abstract

Journal

Organization

Cited Papers