SteerVLM: Robust Model Control through Lightweight Activation Steering for Vision Language Models.

Anushka Sivakumar, Andrew Zhang, Zaber Ibn Abdul Hakim, Christopher Thomas

Browse the full EMNLP paper archive.