Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2023
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2311.09060 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914738190942208 |
|---|---|
| author | Chang, Ting-Yun Thomason, Jesse Jia, Robin |
| author_facet | Chang, Ting-Yun Thomason, Jesse Jia, Robin |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2311_09060 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Do Localization Methods Actually Localize Memorized Data in LLMs? A Tale of Two Benchmarks Chang, Ting-Yun Thomason, Jesse Jia, Robin Computation and Language 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. |
| title | Do Localization Methods Actually Localize Memorized Data in LLMs? A Tale of Two Benchmarks |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2311.09060 |