Erased or Dormant? Rethinking Concept Erasure Through Reversibility
This work addresses the problem of ensuring irreversible concept removal in generative models for AI safety and ethics, but it is incremental as it builds on existing erasure methods.
The paper investigated whether current concept erasure techniques in diffusion models genuinely remove the ability to generate targeted concepts or only achieve superficial suppression, finding that erased concepts often reemerge with substantial visual fidelity after minimal adaptation.
To what extent does concept erasure eliminate generative capacity in diffusion models? While prior evaluations have primarily focused on measuring concept suppression under specific textual prompts, we explore a complementary and fundamental question: do current concept erasure techniques genuinely remove the ability to generate targeted concepts, or do they merely achieve superficial, prompt-specific suppression? We systematically evaluate the robustness and reversibility of two representative concept erasure methods, Unified Concept Editing and Erased Stable Diffusion, by probing their ability to eliminate targeted generative behaviors in text-to-image models. These methods attempt to suppress undesired semantic concepts by modifying internal model parameters, either through targeted attention edits or model-level fine-tuning strategies. To rigorously assess whether these techniques truly erase generative capacity, we propose an instance-level evaluation strategy that employs lightweight fine-tuning to explicitly test the reactivation potential of erased concepts. Through quantitative metrics and qualitative analyses, we show that erased concepts often reemerge with substantial visual fidelity after minimal adaptation, indicating that current methods suppress latent generative representations without fully eliminating them. Our findings reveal critical limitations in existing concept erasure approaches and highlight the need for deeper, representation-level interventions and more rigorous evaluation standards to ensure genuine, irreversible removal of concepts from generative models.