Memory Self-Regeneration: Uncovering Hidden Knowledge in Unlearned Models
This addresses the issue of robust unlearning for AI safety, but it appears incremental as it focuses on evaluating and recovering knowledge rather than introducing a new unlearning method.
The paper tackles the problem of machine unlearning in text-to-image models, where models can still generate harmful content after unlearning, and finds that forgetting occurs in short-term and long-term ways, with recovery being more challenging for long-term forgetting.
The impressive capability of modern text-to-image models to generate realistic visuals has come with a serious drawback: they can be misused to create harmful, deceptive or unlawful content. This has accelerated the push for machine unlearning. This new field seeks to selectively remove specific knowledge from a model's training data without causing a drop in its overall performance. However, it turns out that actually forgetting a given concept is an extremely difficult task. Models exposed to attacks using adversarial prompts show the ability to generate so-called unlearned concepts, which can be not only harmful but also illegal. In this paper, we present considerations regarding the ability of models to forget and recall knowledge, introducing the Memory Self-Regeneration task. Furthermore, we present MemoRa strategy, which we consider to be a regenerative approach supporting the effective recovery of previously lost knowledge. Moreover, we propose that robustness in knowledge retrieval is a crucial yet underexplored evaluation measure for developing more robust and effective unlearning techniques. Finally, we demonstrate that forgetting occurs in two distinct ways: short-term, where concepts can be quickly recalled, and long-term, where recovery is more challenging.