Real World Code
Real-world code research focuses on bridging the gap between large language models (LLMs) and practical software development, aiming to improve the quality, security, and efficiency of automatically generated code. Current research emphasizes developing methods for generating equivalent code representations, ensuring code correctness through techniques like hierarchical debugging and polyhedral modeling, and mitigating security vulnerabilities via prompt optimization and generative adversarial networks. This field is significant because it directly impacts software engineering practices, potentially increasing developer productivity and improving software reliability and security.
Papers
Just another copy and paste? Comparing the security vulnerabilities of ChatGPT generated code and StackOverflow answers
Sivana Hamer, Marcelo d'Amorim, Laurie Williams
Investigating the Performance of Language Models for Completing Code in Functional Programming Languages: a Haskell Case Study
Tim van Dam, Frank van der Heijden, Philippe de Bekker, Berend Nieuwschepen, Marc Otten, Maliheh Izadi
Bugs in Large Language Models Generated Code: An Empirical Study
Florian Tambon, Arghavan Moradi Dakhel, Amin Nikanjam, Foutse Khomh, Michel C. Desmarais, Giuliano Antoniol
Teaching Machines to Code: Smart Contract Translation with LLMs
Rabimba Karanjai, Lei Xu, Weidong Shi
Mastering Text, Code and Math Simultaneously via Fusing Highly Specialized Language Models
Ning Ding, Yulin Chen, Ganqu Cui, Xingtai Lv, Weilin Zhao, Ruobing Xie, Bowen Zhou, Zhiyuan Liu, Maosong Sun
Learn to Code Sustainably: An Empirical Study on LLM-based Green Code Generation
Tina Vartziotis, Ippolyti Dellatolas, George Dasoulas, Maximilian Schmidt, Florian Schneider, Tim Hoffmann, Sotirios Kotsopoulos, Michael Keckeisen
Eliciting Better Multilingual Structured Reasoning from LLMs through Code
Bryan Li, Tamer Alkhouli, Daniele Bonadiman, Nikolaos Pappas, Saab Mansour