Eight Methods to Evaluate Robust Unlearning in LLMs

TL;DR

This paper reviews and applies various evaluation methods to assess the effectiveness of unlearning in large language models, revealing strengths and limitations of current approaches.

Contribution

It provides a comprehensive survey of unlearning evaluation techniques and applies them to a specific model, highlighting the need for standardized, thorough assessment methods.

Findings

WHP's unlearning generalizes well with the 'Familiarity' metric

Higher-than-baseline knowledge can be extracted from WHP

Collateral unlearning occurs in related domains

Abstract

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.