2 papers
cs.LG2025
Leaner Training, Lower Leakage: Revisiting Memorization in LLM Fine-Tuning with LoRA
Fei Wang, Baochun Li
Memorization in large language models (LLMs) makes them vulnerable to data extraction attacks. While pre-training memorization has been extensively studied, fewer works have explor…
cs.LG2025
Hear No Evil: Detecting Gradient Leakage by Malicious Servers in Federated Learning
Fei Wang, Baochun Li
Recent work has shown that gradient updates in federated learning (FL) can unintentionally reveal sensitive information about a client's local data. This risk becomes significantly…