Adaptive Residual-Update Steering for Low-Overhead Hallucination Mitigation in Large Vision Language Models

Open in new window