5 papers
The Poisoned Chalice of LLM Evaluation Report
Jonathan Katzy, Ali Al-Kaswan, Razvan Mihai Popescu +1
Large language models are increasingly used to evaluate and support software engineering tasks, yet the validity of these evaluations is often undermined by uncertainty about wheth…
PrivCode++: Latent-Conditioned Differentially Private Code Generation for Comprehensive Guarantees
Zheng Liu, Chen Gong, Terry Yue Zhuo +6
Large language models fine-tuned on instruction-code pairs may memorize and subsequently leak sensitive training data. Existing differentially private (DP) code generation methods…
LLMs Are Not a Silver Bullet: A Case Study on Software Fairness
Xinyue Li, Sixuan Li, Ying Xiao +4
Fairness is a critical requirement for human-related, high-stakes software systems, motivating extensive research on bias mitigation. Prior work has largely focused on tabular data…
AEGIS: From Clues to Verdicts -- Graph-Guided Deep Vulnerability Reasoning via Dialectics and Meta-Auditing
Sen Fang, Weiyuan Ding, Zhezhen Cao +2
Large Language Models (LLMs) are increasingly adopted for vulnerability detection, yet their reasoning remains fundamentally unsound. We identify a root cause shared by both major…
How Quantization Impacts Privacy Risk on LLMs for Code?
Md Nazmul Haque, Hua Yang, Zhou Yang +1
Large language models for code (LLMs4Code) rely heavily on massive training data, including sensitive data, such as cloud service credentials of the projects and personal identifia…