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

2025/05/15 by Ningxin Gui, Gui, Ningxin, Jia, Qianghuai +7 · 1 citation
Computer Science · Materials Science · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning in Materials Science #Natural Language Processing Techniques #Software Engineering (cs.SE) #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2505.10594

openalex publication_date 2025/05/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

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.

Citations

• o1-Coder: an o1 Replication for Coding
• Process Supervision-Guided Policy Optimization for Code Generation
• CodeDPO: Aligning Code Models with Self Generated and Verified Source Code
• LLaMA-Berry: Pairwise Optimization for O1-like Olympiad-Level Mathematical Reasoning
• Qwen2.5-Coder Technical Report
• DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search
• Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMs
• DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence
• AlphaMath Almost Zero: Process Supervision without Process
• StarCoder 2 and The Stack v2: The Next Generation
• StepCoder: Improve Code Generation with Reinforcement Learning from Compiler Feedback
• DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence
• WaveCoder: Widespread And Versatile Enhancement For Code Large Language Models By Instruction Tuning
• Magicoder: Empowering Code Generation with OSS-Instruct
• Code Llama: Open Foundation Models for Code
• RLTF: Reinforcement Learning from Unit Test Feedback
• WizardCoder: Empowering Code Large Language Models with Evol-Instruct
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model
• Tree of Thoughts: Deliberate Problem Solving with Large Language Models
• Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generation
• CodeRL: Mastering Code Generation through Pretrained Models and Deep Reinforcement Learning
• Compilable Neural Code Generation with Compiler Feedback
• BNAI, NO-TOKEN, and MIND-UNITY: Pillars of a Systemic Revolution in Artificial Intelligence

Cited by

• Reinforcement Learning-Guided Chain-of-Draft for Token-Efficient Code Generation

Related

• Evaluating Code Reasoning Abilities of Large Language Models Under Real-World Settings
• Seed-Coder: Let the Code Model Curate Data for Itself
• CompCodeVet: A Compiler-guided Validation and Enhancement Approach for Code Dataset
• From Reasoning to Code: GRPO Optimization for Underrepresented Languages
• SEER: Enhancing Chain-of-Thought Code Generation through Self-Exploring Deep Reasoning