Sahar Abdelnabi

LLMail-Inject: A Dataset from a Realistic Adaptive Prompt Injection Challenge

Terrarium: Revisiting the Blackboard for Studying Multi-Agent Attacks

Contextual Integrity in LLMs via Reasoning and Reinforcement Learning

Can LLMs Separate Instructions From Data? And What Do We Even Mean By That?

The Hawthorne Effect in Reasoning Models: Evaluating and Steering Test Awareness

Filling the Gaps: LLMs for Causal Hypothesis Generation

Theory of LLM sampling: part descriptive and part prescriptive