AI RESEARCH
Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection
arXiv CS.LG
•
ArXi:2605.28664v1 Announce Type: new Safety detection models require examples of HHH (Helpful, Harmless, Honest)-violating outputs for robust generalization, however such examples are scarce. Activation Steering (AS) has emerged as a data-efficient method for generating target-concept-aligned responses. We investigate whether AS can generate high-quality