Poster · Sydney
Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models
- When
- Tue 8 Dec 2026, 10:00–13:00 AEDT
- Where
- Hall 1-4, ICC Sydney
Exhibition Centre, Level 1 - Session
- Sydney Poster Session 1
- Authors
- Chen Xiong, Zhiyuan HE, Pin-Yu Chen, Ching-Yun Ko, Tsung-Yi Ho
- Organisations
- The Chinese University of Hong Kong, IBM
- QRS classification
- Language models, Safety, fairness and privacy