UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation. PDF.Link.

Jiyu Guo, Shuo Yang, Yiming Huang, Yancheng Long, Xiaobo Xia, Xiu Su, Bo Zhao, Zeke Xie, Liqiang Nie.

• Trainging Data Genaration.
• In this paper, we pioneer a paradigm shift from intrinsic data quality to task-specific utility optimization in generative data augmentation

Robust Coreset Selection via Class-Aware Decision Boundary Reconstruction.

Shuo Yang, Jiyu Guo, Yujie Wei, Ruiheng Zhang, Hongxun Yao, Ping Luo, Tongliang Liu, Liqiang Nie(Under Review, Student First Author)

• Training Data Selection.
• We establish a principled coreset selection method by linking decision boundary reconstruction error to model generalization, enabling robust and efficient training data selection via decision boundary reconstruction.

Can LLMs Replace Human Evaluators? An Empirical Study of LLM-as-a-Judge in Software Engineering.PDF.

Ruiqi Wang, Jiyu Guo, Cuiyun Gao, Guodong Fan, Chun Yong Chong, Xin Xia.

• LLM-as-a-judge.
• In this paper, we empirically explore LLM-as-a-judge methods for evaluating SE tasks, focusing on their alignment with human judgments.

Enhancing text-to-SQL parsing through question rewriting and execution-guided refinement.PDF.

Wenxin Mao, Ruiqi Wang, Jiyu Guo, Jichuan Zeng, Cuiyun Gao, Peiyi Han, Chuanyi Liu.

• AI for Databases.
• We build a modular, execution-driven Text-to-SQL agent that enhances semantic alignment and query correctness via context-aware reasoning and iterative self-correction.