SemCoder: Training Code Language Models with Comprehensive Semantics Reasoning

Yangruibo Ding, Jinjun Peng, Marcus J. Min +3

Code Large Language Models (Code LLMs) have excelled at tasks like code completion but often miss deeper semantics such as execution effects and dynamic states. This paper aims to…