Simon Lermen

1. Large-scale online deanonymization with LLMs
   LLMs enable scalable, high-precision deanonymization of pseudonymous online accounts using unstructured text.
   2026/02/18 by Simon Lermen, Daniel Paleka, Joshua Swanson +3 · 85 voices · 1 citation
   #cs.CR #cs.AI #cs.LG
2. LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70B
   2023/10/31 by Simon Lermen, Charlie Rogers-Smith, Lermen, Simon +4 · 2 voices · 31 citations
   Computer Science · Medicine · #Adversarial Robustness in Machine Learning #Artificial Intelligence in Healthcare and Education #cs.AI #cs.LG
3. Evaluating Large Language Models' Capability to Launch Fully Automated Spear Phishing Campaigns: Validated on Human Subjects
   2024/11/30 by Fred Heiding, Heiding, Fred, Simon Lermen +7 · 5 voices · 4 citations
   Computer Science · Social Sciences · #Spam and Phishing Detection #Misinformation and Its Impacts #Topic Modeling
4. BadLlama: cheaply removing safety fine-tuning from Llama 2-Chat 13B
   2023/10/31 by P.R. Gade, Gade, Pranav, Simon Lermen +5 · 3 citations
   Computer Science · #Hate Speech and Cyberbullying Detection #Adversarial Robustness in Machine Learning
5. Deceptive Automated Interpretability: Language Models Coordinating to Fool Oversight Systems
   2025/04/10 by Simon Lermen, Mateusz Dziemian, Lermen, Simon +3 · 1 voice · 2 citations
   #cs.AI #cs.CL