AI Safety Research Exchange

​Join fellow AI safety researchers for a collaborative gathering focused on sharing current work and exploring new opportunities.

​At this Research Exchange, participants will deliver brief lightning talks highlighting their ongoing projects, research directions they’re eager to explore, and areas where they’re seeking collaboration.

​Whether you’re looking to find research collaborators, get feedback on early-stage ideas, or simply stay informed about developments across the AI safety landscape, this Research Exchange offers a focused environment to build connections that advance this important field.

​If you are interested to present, please indicate in the response form and we will reach out to you.

​Lightning talks:

• ​Confidence elicitation: A new attack vector for LLMs

• ​Reasoning LLMs are Wandering Solution Explorers

• ​Bullying the Machine: How Personas Increase LLM Vulnerability

• ​Mechanistic Interpretability - Interpreting refusal in LLMs