Authors: Alberto Blanco-Justicia, Najeeb Jebreel, Benet Manzanares, David Sánchez, Josep Domingo-Ferrer, Guillem Collell, Kuan Eeik Tan

Published on: April 02, 2024

Impact Score: 8.4

Arxiv code: Arxiv:2404.02062

Summary

• What is new: A detailed survey and taxonomy of digital forgetting in large language models (LLMs), highlighting unlearning methods as the state of the art.
• Why this is important: The need to effectively remove undesirable knowledge or behavior from models while retaining their performance on desired tasks.
• What the research proposes: Utilizing unlearning methodologies for efficient and scalable digital forgetting in LLMs.
• Results: A comprehensive comparison of current approaches, and an evaluation of datasets, models, and metrics for assessing forgetting, retaining, and runtime.

Technical Details

Technological frameworks used: Machine unlearning methods specific to LLMs are discussed, including their types and applications.

Models used: Various large language models are surveyed and compared based on their ability to forget and retain information.

Data used: Datasets and metrics for evaluating the effectiveness of digital forgetting strategies are detailed.

Potential Impact

Privacy protection, online content creation, AI ethics organizations, and any entity relying on large language models for data processing or generation.

Want to implement this idea in a business?

We have generated a startup concept here: ForgetMeNot Technologies.