Chris Potts

3 papers Stanford University

• Blackbox Model Provenance via Palimpsestic Membership Inference
• Do Language Models Use Their Depth Efficiently?
• Improved Representation Steering for Language Models