Trying to Steer PI with (Prefix) Label Directions
TL;DR Building on the method in Prompt Injection as Role Confusion, we prepend earlier label blocks, with empty contents, to the neutral-text block used to train the probe. We compare this prefix method with the original method (i.e., bare) for both probing and steering, across three models—GPT-OSS-20B, Qwen3.5-9B, and Gemma4-12B—and...
Nov 81