Digital Forgetting in Large Language Models: A Survey of Unlearning Methods
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909454735245312 |
|---|---|
| author | Blanco-Justicia, Alberto Jebreel, Najeeb Manzanares, Benet Sánchez, David Domingo-Ferrer, Josep Collell, Guillem Tan, Kuan Eeik |
| author_facet | Blanco-Justicia, Alberto Jebreel, Najeeb Manzanares, Benet Sánchez, David Domingo-Ferrer, Josep Collell, Guillem Tan, Kuan Eeik |
| contents | The objective of digital forgetting is, given a model with undesirable knowledge or behavior, obtain a new model where the detected issues are no longer present. The motivations for forgetting include privacy protection, copyright protection, elimination of biases and discrimination, and prevention of harmful content generation. Effective digital forgetting has to be effective (meaning how well the new model has forgotten the undesired knowledge/behavior), retain the performance of the original model on the desirable tasks, and be scalable (in particular forgetting has to be more efficient than retraining from scratch on just the tasks/data to be retained). This survey focuses on forgetting in large language models (LLMs). We first provide background on LLMs, including their components, the types of LLMs, and their usual training pipeline. Second, we describe the motivations, types, and desired properties of digital forgetting. Third, we introduce the approaches to digital forgetting in LLMs, among which unlearning methodologies stand out as the state of the art. Fourth, we provide a detailed taxonomy of machine unlearning methods for LLMs, and we survey and compare current approaches. Fifth, we detail datasets, models and metrics used for the evaluation of forgetting, retaining and runtime. Sixth, we discuss challenges in the area. Finally, we provide some concluding remarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_02062 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Digital Forgetting in Large Language Models: A Survey of Unlearning Methods Blanco-Justicia, Alberto Jebreel, Najeeb Manzanares, Benet Sánchez, David Domingo-Ferrer, Josep Collell, Guillem Tan, Kuan Eeik Cryptography and Security Artificial Intelligence Machine Learning 68 K.4.1; I.2.6; I.2.7 The objective of digital forgetting is, given a model with undesirable knowledge or behavior, obtain a new model where the detected issues are no longer present. The motivations for forgetting include privacy protection, copyright protection, elimination of biases and discrimination, and prevention of harmful content generation. Effective digital forgetting has to be effective (meaning how well the new model has forgotten the undesired knowledge/behavior), retain the performance of the original model on the desirable tasks, and be scalable (in particular forgetting has to be more efficient than retraining from scratch on just the tasks/data to be retained). This survey focuses on forgetting in large language models (LLMs). We first provide background on LLMs, including their components, the types of LLMs, and their usual training pipeline. Second, we describe the motivations, types, and desired properties of digital forgetting. Third, we introduce the approaches to digital forgetting in LLMs, among which unlearning methodologies stand out as the state of the art. Fourth, we provide a detailed taxonomy of machine unlearning methods for LLMs, and we survey and compare current approaches. Fifth, we detail datasets, models and metrics used for the evaluation of forgetting, retaining and runtime. Sixth, we discuss challenges in the area. Finally, we provide some concluding remarks. |
| title | Digital Forgetting in Large Language Models: A Survey of Unlearning Methods |
| topic | Cryptography and Security Artificial Intelligence Machine Learning 68 K.4.1; I.2.6; I.2.7 |
| url | https://arxiv.org/abs/2404.02062 |