Method helps AI forget specific knowledge and admit what it does not know
Decoupling Knowledge Forgetting and Refusal Learning in MLLM Unlearning
Artificial IntelligenceComputation and LanguageInformation RetrievalMachine Learning
Summary
AI models sometimes need to forget certain information, but simply blocking answers can make them guess wrong or hide knowledge without truly forgetting it. The authors created a method called Forget-Aware Self-Distillation that helps AI not only lose specific facts but also honestly say it doesn’t know them. This approach uses two parts: one that teaches the AI to refuse to answer and another that removes the actual knowledge inside it. Their tests show the AI forgets correctly and avoids lying or guessing when asked about the forgotten topics.
What this means in practice
- •For ai application developers: Build AI systems that can reliably forget sensitive or outdated information while honestly refusing to answer related queries.
- •For data privacy teams: Improve compliance by ensuring AI models genuinely erase specified user data rather than merely hiding it from responses.
Authors
Haodong Chen, Shengtao Wen, Dong Liang, Lingbing Guo, Xiang Chen
Abstract
Current evaluations of multimodal unlearning typically focus on whether a model stops outputting target information. This surface-level criterion, however, masks two critical failure modes. First, knowledge suppression techniques often induce confabulation, leading the model to hallucinate facts about the unlearned subject. Second, refusal training merely alters observable behavior without erasing the underlying knowledge; such models might decline open queries but still identify correct answers under multiple-choice probing. To overcome these limitations, we propose Forget-Aware Self-Distillation (FASD). This method is designed to achieve honest ignorance, a state where the model genuinely loses the target knowledge and learns to acknowledge its absence. FASD separates these goals into two complementary channels. The behavior channel distills appropriate refusals from an ignorance-prompted copy of the model itself. Crucially, target tokens are masked from the teacher distribution to avoid accidentally reteaching the forgotten concepts. Simultaneously, the knowledge channel directly suppresses ground-truth answer spans independently of the model's generated responses. Experiments demonstrate that FASD successfully unlearns target knowledge, prevents information leakage in open-ended tasks, and preserves performance on retained data. The resulting model not only forgets the specified information but accurately recognizes the boundaries of its own knowledge. Code and data will be released upon publication.