{"type":"ScholarlyArticle","title":"ProCURE: Addressing the Programming Concept Understanding Gap for Code Generation in LLMs via Concept-Aware Consistency Learning","authors":[{"name":"Xiaoning Ren","slug":"xiaoning-ren","given":"Xiaoning","family":"Ren","affiliation":""},{"name":"Qiang Hu","slug":"qiang-hu","given":"Qiang","family":"Hu","affiliation":""},{"name":"Wei Ma","slug":"wei-ma","given":"Wei","family":"Ma","affiliation":""},{"name":"Chongyang Liu","slug":"chongyang-liu","given":"Chongyang","family":"Liu","affiliation":""},{"name":"Yan Li","slug":"yan-li","given":"Yan","family":"Li","affiliation":""},{"name":"Yao Zhang","slug":"yao-zhang","given":"Yao","family":"Zhang","affiliation":""},{"name":"Lingxiao Jiang","slug":"lingxiao-jiang","given":"Lingxiao","family":"Jiang","affiliation":""},{"name":"Yongqiang Lyu","slug":"yongqiang-lyu","given":"Yongqiang","family":"Lyu","affiliation":""},{"name":"Yinxing Xue","slug":"yinxing-xue","given":"Yinxing","family":"Xue","affiliation":""}],"abstract":"Although Large Language Models (LLMs) excel at code generation, recent research reveals that they exhibit an insufficient grasp of core programming concepts, such as data flow and control flow. This limitation undermines their robustness when encountering variations in these concepts in practice; however, effective solutions that explicitly target this gap remain limited.\r\nTo address this challenge, we propose  ProCURE, a concept-aware consistency learning framework designed to enhance LLMs’ understanding of programming concepts. Specifically, ProCURE first performs automated concept-oriented code augmentation to construct a concept-aligned dataset covering  representative programming concepts. It then conducts concept-aware  fine-tuning, encouraging the model to capture fine-grained concept variations and learn appropriate generation behaviors under such variations via a novel concept-sensitive consistency loss.\r\nTo quantify programming concept understanding, we introduce the Concept Consistency Score (CCScore), defined as the proportion of correct generations preserved under concept variations.  A higher CCScore indicates a more profound understanding of programming concepts.\r\nWe evaluate ProCURE on four open-source LLMs across three widely used code generation benchmarks.  Experimental results show that ProCURE improves CCScore by an average of 17.9 points, demonstrating its effectiveness in addressing the programming concept understanding gap.","doi":"10.24963/ijcai.2026/529","url":"https://ijcai.org/proceedings/2026/529","pdf":"https://ijcai.org/proceedings/2026/0529.pdf","markdown":"https://ijcai.org/proceedings/2026/529.md","bibtex":"https://ijcai.org/proceedings/2026/bibtex/529","volume":{"slug":"2026","title":"Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2026","url":"https://ijcai.org/proceedings/2026/","year":2026,"location":"Bremen, Germany","dates":"15-21 August 2026","editors":"Diego Calvanese","isbn":"978-1-956792-09-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2026"},"section":"Machine Learning","track":"Main Track","number":529,"pages":{"first":4750,"last":4758},"keywords":["Machine Learning: Applications","Natural Language Processing: Applications"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}