Probabilistic Concept-Aware Steering for Trustworthy LLM Inference

Researchers propose the Probabilistic Concept-Aware Steering (PCS) framework for large language models (LLMs), which uses concept-driven steering-vector retrieval and probabilistic strength calibration to provide controllable, safety-oriented semantic bias while preserving original task competence. The framework addresses representation-incoherent behaviors in existing steering vector methods by moving beyond binary positive-negative evaluation and discrete clustering metrics.

arXiv:2607.18259v1 Announce Type: new Abstract: Steering vectors SVs , an inference-time intervention technique for large language models LLMs , guide the generation process by adding a concept-specific direction vector to intermediate activations during inference. However, existing SV methods frequently yield representation-incoherent behaviors that undermine interpretability and fine-grained control, largely because prior work has focused on binary positive-negative steering evaluation while employing discrete clustering metrics that fail to capture the continuous spectrum of semantic alignment. In this work, we present the Probabilistic Concept-Aware Steering PCS framework for LLM inference. PCS preserves original task competence while providing controllable, safety-oriented semantic bias through concept-driven steering-vector retrieval and probabilistic strength calibration.