כיווני תופעה רדומה בחלל סמוי של LLMs חותרים לבטיחות, ולא לצדק

תקציר מקורי באנגליתarXiv:2610.08559v1 Announce Type: new Abstract: Activation steering has gained popularity as a lightweight inference-time debiasing technique for large language models. However, prior work reports that steering vectors generalise poorly, with unintended effects on model performance and limited transfer to new datasets. Our work analyses what the debiasing direction used for activation steering actually encodes, in order to shed light on its inconsistent performance. We study the linear debiasing direction obtained by contrasting the activations of anti-biased and biased prompts, and evaluate it as a steering intervention across bias and general knowledge benchmarks. We find that this direction is dominated by model confidence, pointing from regions of high to low-probability tokens in acti