Eight Methods to Evaluate Robust Unlearning in LLMs
This work addresses the problem of rigorously assessing unlearning for researchers and practitioners, but it is incremental as it builds on existing models and metrics.
The paper tackles the lack of standardized methods for evaluating machine unlearning in LLMs by surveying existing techniques and applying eight tests to the 'Who's Harry Potter' model, finding that while it generalizes well with one metric, it retains higher-than-baseline knowledge, performs similarly on tasks, and shows collateral unlearning.
Machine unlearning can be useful for removing harmful capabilities and memorized text from large language models (LLMs), but there are not yet standardized methods for rigorously evaluating it. In this paper, we first survey techniques and limitations of existing unlearning evaluations. Second, we apply a comprehensive set of tests for the robustness and competitiveness of unlearning in the "Who's Harry Potter" (WHP) model from Eldan and Russinovich (2023). While WHP's unlearning generalizes well when evaluated with the "Familiarity" metric from Eldan and Russinovich, we find i) higher-than-baseline amounts of knowledge can reliably be extracted, ii) WHP performs on par with the original model on Harry Potter Q&A tasks, iii) it represents latent knowledge comparably to the original model, and iv) there is collateral unlearning in related domains. Overall, our results highlight the importance of comprehensive unlearning evaluation that avoids ad-hoc metrics.