Mechanistic Interpretability for Steering Vision-Language-Action Models

Paper: https://arxiv.org/pdf/2509.00328