Selected work

Representative Papers

AI-Driven Grading and Moderation for Collaborative Projects in Computer Science Education

Oct 04, 2025

To address the challenge of fairly, objectively, and scalably assessing individual contributions in collaborative programming projects within computer science education, this paper proposes an AI-assisted semi-automated grading framework. Methodologically, it integrates heterogeneous behavioral signals—including Git repository activities (commits, code modifications, code reviews), GitHub communication logs (issue comments, pull request interactions), and temporal collaboration patterns—leveraging natural language processing and supervised learning to construct an interpretable, multi-source model for quantifying individual contributions. The key innovation lies in the first unified modeling of code evolution, socio-technical interactions, and temporal dynamics, augmented with pedagogically transparent attribution mechanisms. Evaluated in an upper-level undergraduate course, the system achieves a correlation of 0.92 (p < 0.01) with instructor assessments, increases student satisfaction by 27%, and reduces instructor grading time by 68%.

0 citationsRead paper

AutoAdv: Automated Adversarial Prompting for Multi-Turn Jailbreaking of Large Language Models

Apr 18, 2025arXiv.org

This work addresses the security vulnerability of large language models (LLMs) to jailbreaking attacks. Methodologically, it proposes a dynamic, multi-round adversarial prompt generation framework that integrates role-playing, context manipulation, and semantic obfuscation into a parameterized attack model. Guided by failure analysis, the framework iteratively refines prompts across dialogue turns and enhances attack efficacy via strategic system prompt engineering and hyperparameter optimization. Its key contribution is the first automated, interpretable, and high-success-rate jailbreaking method tailored to complex, multi-turn conversational scenarios. Experiments on mainstream LLMs—including ChatGPT, Llama-3, and DeepSeek-V2—achieve an 86% jailbreaking success rate, revealing critical structural weaknesses in current safety mechanisms under multi-turn interaction. The approach establishes a novel paradigm for LLM red-teaming evaluation.

0 citationsRead paper

Recent publications

Latest Papers

AI-Driven Grading and Moderation for Collaborative Projects in Computer Science Education

Oct 04, 2025

To address the challenge of fairly, objectively, and scalably assessing individual contributions in collaborative programming projects within computer science education, this paper proposes an AI-assisted semi-automated grading framework. Methodologically, it integrates heterogeneous behavioral signals—including Git repository activities (commits, code modifications, code reviews), GitHub communication logs (issue comments, pull request interactions), and temporal collaboration patterns—leveraging natural language processing and supervised learning to construct an interpretable, multi-source model for quantifying individual contributions. The key innovation lies in the first unified modeling of code evolution, socio-technical interactions, and temporal dynamics, augmented with pedagogically transparent attribution mechanisms. Evaluated in an upper-level undergraduate course, the system achieves a correlation of 0.92 (p < 0.01) with instructor assessments, increases student satisfaction by 27%, and reduces instructor grading time by 68%.

0 citationsRead paper

AutoAdv: Automated Adversarial Prompting for Multi-Turn Jailbreaking of Large Language Models

Apr 18, 2025arXiv.org

This work addresses the security vulnerability of large language models (LLMs) to jailbreaking attacks. Methodologically, it proposes a dynamic, multi-round adversarial prompt generation framework that integrates role-playing, context manipulation, and semantic obfuscation into a parameterized attack model. Guided by failure analysis, the framework iteratively refines prompts across dialogue turns and enhances attack efficacy via strategic system prompt engineering and hyperparameter optimization. Its key contribution is the first automated, interpretable, and high-success-rate jailbreaking method tailored to complex, multi-turn conversational scenarios. Experiments on mainstream LLMs—including ChatGPT, Llama-3, and DeepSeek-V2—achieve an 86% jailbreaking success rate, revealing critical structural weaknesses in current safety mechanisms under multi-turn interaction. The approach establishes a novel paradigm for LLM red-teaming evaluation.

0 citationsRead paper