CRPE: Expanding The Reasoning Capability of Large Language Model for Code Generation

2025-05-15

AI Summary

Large language models (LLMs) exhibit limited logical reasoning and analytical capabilities in code generation tasks. Method: This paper proposes CRPE, a three-stage framework built upon Qwen2.5-Coder, integrating chain-of-thought (CoT) modeling, StepDPO reinforcement learning, and multi-stage synthetic data distillation. It introduces a novel, code-reasoning–oriented autonomous data synthesis and closed-loop enhancement mechanism, enabling an end-to-end open-source pipeline—from instruction parsing and expert-level reasoning data construction to iterative model capability refinement. Contribution/Results: Experiments demonstrate that COT-Coder-7B-StepDPO achieves a pass@1 score of 21.88 on LiveCodeBench, outperforming same-scale models; COT-Coder-32B-StepDPO attains 35.08, surpassing GPT-4o. Both variants significantly advance code-level logical reasoning performance.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

We introduce CRPE (Code Reasoning Process Enhancer), an innovative three-stage framework for data synthesis and model training that advances the development of sophisticated code reasoning capabilities in large language models (LLMs). Building upon existing system-1 models, CRPE addresses the fundamental challenge of enhancing LLMs' analytical and logical processing in code generation tasks. Our framework presents a methodologically rigorous yet implementable approach to cultivating advanced code reasoning abilities in language models. Through the implementation of CRPE, we successfully develop an enhanced COT-Coder that demonstrates marked improvements in code generation tasks. Evaluation results on LiveCodeBench (20240701-20240901) demonstrate that our COT-Coder-7B-StepDPO, derived from Qwen2.5-Coder-7B-Base, with a pass@1 accuracy of 21.88, exceeds all models with similar or even larger sizes. Furthermore, our COT-Coder-32B-StepDPO, based on Qwen2.5-Coder-32B-Base, exhibits superior performance with a pass@1 accuracy of 35.08, outperforming GPT4O on the benchmark. Overall, CRPE represents a comprehensive, open-source method that encompasses the complete pipeline from instruction data acquisition through expert code reasoning data synthesis, culminating in an autonomous reasoning enhancement mechanism.

Problem

• Enhancing LLMs' analytical and logical processing in code generation
• Developing advanced code reasoning abilities in language models
• Improving pass@1 accuracy in code generation benchmarks

Innovation

• Three-stage framework for code reasoning enhancement
• Advanced COT-Coder with improved code generation
• Open-source pipeline from data to reasoning mechanism