Erasing More Than Intended? How Concept Erasure Degrades the Generation of Non-Target Concepts

Documenti analoghi

• What Secrets Do Your Manifolds Hold? Understanding the Local Geometry of Generative Models
  di: Humayun, Ahmed Imtiaz, et al.
  Pubblicazione: (2024)
• Position: Cracking the Code of Cascading Disparity Towards Marginalized Communities
  di: Farnadi, Golnoosh, et al.
  Pubblicazione: (2024)
• Delta-Crosscoder: Robust Crosscoder Model Diffing in Narrow Fine-Tuning Regimes
  di: Kassem, Aly, et al.
  Pubblicazione: (2026)
• Neighbor-Aware Localized Concept Erasure in Text-to-Image Diffusion Models
  di: Shi, Zhuan, et al.
  Pubblicazione: (2026)
• Reviving Your MNEME: Predicting The Side Effects of LLM Unlearning and Fine-Tuning via Sparse Model Diffing
  di: Kassem, Aly M., et al.
  Pubblicazione: (2025)