Poster · Atlanta

A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models

Name: A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models
Start: 2026-12-09T16:30:00-05:00
End: 2026-12-09T19:30:00-05:00
Location: Hall C1, Georgia World Congress Center

When: Wed 9 Dec 2026, 16:30–19:30 EST
Where: Hall C1, Georgia World Congress Center
Session: Atlanta Poster Session 2
Authors: Hamid Kazemi, Atoosa Malemir Chegini, Maria Safi
Organisations: Apple, University of Maryland, College Park
QRS classification: Language models, Safety, fairness and privacy

Open in schedule Official page