Lifecycle-Aware code generation: Leveraging Software Engineering Phases in LLMs

2025-10-27

AI Summary

Contemporary large language models predominantly adopt a “requirement-to-code” one-step generation paradigm, overlooking structured intermediate artifacts inherent in software engineering. Method: This paper proposes a lifecycle-aware code generation framework that, for the first time, systematically incorporates software development lifecycle (SDLC) intermediates—such as requirement analysis, state machine modeling, and pseudocode—as explicit reasoning and training scaffolds. Our approach integrates multi-stage reasoning with lifecycle-level fine-tuning to explicitly model the software development process. Contribution/Results: On DeepSeek-Coder-1.3B, our method achieves up to a 75% absolute improvement in functional correctness and attains a CodeBLEU score of 34.3% (a relative gain of 11.2%). Notably, it maintains robust performance using ≤80% of the original training data. Ablation studies confirm that state machine modeling is the most critical component, delivering the largest marginal contribution to overall performance.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent progress in large language models (LLMs) has advanced automatic code generation, yet most approaches rely on direct, single-step translation from problem descriptions to code, disregarding structured software engineering practices. We introduce a lifecycle-aware framework that systematically incorporates intermediate artifacts such as requirements analysis, state machine modeling, and pseudocode into both the training and inference stages. This design aligns code generation with standard software development phases and enables more structured reasoning. Experiments show that lifecycle-level fine-tuning improves code correctness by up to 75% over the same model before fine-tuning, with performance gains compounding across intermediate stages. Multi-step inference consistently surpasses single-step generation, demonstrating the effectiveness of intermediate scaffolding. Notably, open-source LLMs, once fine-tuned under our framework, match or slightly outperform models pretrained on code. When applied to DeepSeek-Coder-1.3B, our framework yields relative CodeBLEU improvements of 34.3%, 20.0%, 11.2%, and 22.3% over ChatGPT-3.5, ChatGPT-4o-mini, DeepSeek-R1, and LLaMA-8B, respectively. Our pipeline also proves robust with up to 80% less training data, confirming its resilience. Ablation studies further reveal that each intermediate artifact contributes distinctly to final code quality, with state machine modeling yielding the most substantial impact. Our source code and detailed experimental data are available at https://anonymous.4open.science/r/Lifecycle-Aware-3CCB.

Problem

• Improving code generation by incorporating software engineering phases
• Enhancing code correctness through lifecycle-aware fine-tuning approach
• Integrating intermediate artifacts like requirements analysis and pseudocode

Innovation

• Integrates software engineering phases into code generation
• Uses lifecycle-level fine-tuning to improve code correctness
• Employs multi-step inference with intermediate artifacts