Code Model
Code models, large language models (LLMs) trained on vast code datasets, aim to automate various software engineering tasks, such as code generation, debugging, and understanding. Current research focuses on improving model accuracy and efficiency through techniques like synthetic data generation (e.g., using code edits or program diffs), reinforcement learning for performance optimization, and contrastive learning for robustness. These advancements are significant because they promise to increase programmer productivity, improve code quality and security, and enable new applications in software development and beyond.
Papers
ML-Bench: Evaluating Large Language Models and Agents for Machine Learning Tasks on Repository-Level Code
Xiangru Tang, Yuliang Liu, Zefan Cai, Yanjun Shao, Junjie Lu, Yichi Zhang, Zexuan Deng, Helan Hu, Kaikai An, Ruijun Huang, Shuzheng Si, Sheng Chen, Haozhe Zhao, Liang Chen, Yan Wang, Tianyu Liu, Zhiwei Jiang, Baobao Chang, Yin Fang, Yujia Qin, Wangchunshu Zhou, Yilun Zhao, Arman Cohan, Mark Gerstein
Code Models are Zero-shot Precondition Reasoners
Lajanugen Logeswaran, Sungryull Sohn, Yiwei Lyu, Anthony Zhe Liu, Dong-Ki Kim, Dongsub Shim, Moontae Lee, Honglak Lee