Abstract
Modular programming, which aims to construct the final program by integrating smaller, independent building blocks, has been regarded as a desirable practice in software development. However, with the rise of recent code generation agents built upon large language models (LLMs), a question emerges: is this traditional practice equally effective for these new tools? In this work, we assess the impact of modularity in code generation by introducing a novel metric for its quantitative measurement. Surprisingly, unlike conventional wisdom on the topic, we find that modularity is not a core factor for improving the performance of code generation models. We also explore potential explanations for why LLMs do not exhibit a preference for modular code compared to non-modular code.- Anthology ID:
- 2024.findings-emnlp.676
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2024
- Month:
- November
- Year:
- 2024
- Address:
- Miami, Florida, USA
- Editors:
- Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 11561–11571
- Language:
- URL:
- https://aclanthology.org/2024.findings-emnlp.676
- DOI:
- 10.18653/v1/2024.findings-emnlp.676
- Cite (ACL):
- Deokyeong Kang, KiJung Seo, and Taeuk Kim. 2024. Revisiting the Impact of Pursuing Modularity for Code Generation. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 11561–11571, Miami, Florida, USA. Association for Computational Linguistics.
- Cite (Informal):
- Revisiting the Impact of Pursuing Modularity for Code Generation (Kang et al., Findings 2024)
- PDF:
- https://preview.aclanthology.org/dois-2013-emnlp/2024.findings-emnlp.676.pdf