Improving Code Understanding in Large Language Models through Concept-Aware Consistency Learning
For developers using LLMs for code tasks, this work addresses the shallow conceptual understanding that limits practical adoption, but the improvement is incremental.
LLMs struggle with fundamental programming concepts like data flow and control flow, leading to fragile performance on code tasks. The authors propose a counterfactual code augmentation framework with concept-aware tuning that improves LLMs' conceptual understanding, achieving stronger results across multiple models and benchmarks.
Large language models (LLMs) have recently shown impressive results on diverse code-related tasks, benefiting from large-scale training and instruction tuning. However, studies reveal that their grasp of fundamental programming concepts, such as data flow and control flow, remains shallow, leading to fragile performance when code requires deeper reasoning. This limitation restricts the practical adoption of LLMs in real-world software development. To address this issue, this work introduces a counterfactual code augmentation framework combined with concept-aware tuning, designed to guide LLMs toward stronger conceptual understanding. Comprehensive evaluation across multiple models and benchmarks demonstrates the effectiveness of the proposed approach.