Do Localization Methods Actually Localize Memorized Data in LLMs? A Tale of Two Benchmarks
This work addresses the need for systematic evaluation of localization methods in LLMs, which is incremental as it builds on prior concepts but provides new benchmarks.
The authors tackled the problem of evaluating localization methods in LLMs by proposing two benchmarks (INJ and DEL) to test if these methods can pinpoint components responsible for memorized data, finding that methods adapted from network pruning performed well and all showed promising ability, though none identified sequence-specific neurons.
The concept of localization in LLMs is often mentioned in prior work; however, methods for localization have never been systematically and directly evaluated. We propose two complementary benchmarks that evaluate the ability of localization methods to pinpoint LLM components responsible for memorized data. In our INJ benchmark, we actively inject a piece of new information into a small subset of LLM weights, enabling us to directly evaluate whether localization methods can identify these "ground truth" weights. In our DEL benchmark, we evaluate localization by measuring how much dropping out identified neurons deletes a memorized pretrained sequence. Despite their different perspectives, our two benchmarks yield consistent rankings of five localization methods. Methods adapted from network pruning perform well on both benchmarks, and all evaluated methods show promising localization ability. On the other hand, even successful methods identify neurons that are not specific to a single memorized sequence.