Shauli Ravfogel

Latest

• A Practical Method for Generating String Counterfactuals
• Gumbel Counterfactual Generation from Language Models
• On Affine Homotopy between Language Encoders
• Representation Surgery: Theory and Practice of Affine Steering
• Log-Linear Guardedness and Its Implications
• LEACE: Perfect linear concept erasure in closed form
• Kernelized Concept Erasure
• Linear Adversarial Concept Erasure
• Visual Comparison of Language Model Adaptation