Tag: Pre-trained models

Create an illustration in a vibrant and cheerful style in a 3:2 aspect ratio. The image should vividly describe the core concept of the article. It should represent a new learning paradigm - Non-fine-tunability, designed to prevent misuse of pre-trained models for inappropriate tasks. Show a fortified labyrinth containing a brain-like computer model representing the SOPHON protection framework. The labyrinth's complexity signifies overhead comparable to or greater than training from scratch when fine-tuning in restricted domains, thus resisting misuse of this model. This illustration should be filled with light colors to indicate positivity and bright future prospects of this technology.

Sophon: Non-Fine-Tunable Learning to Restrain Task Transferability For Pre-trained Models

The research paper introduces a new learning paradigm called non-fine-tunable learning, which aims to prevent…
Read More
Deep learning Non-fine-tunable learning Pre-trained models Sophon Transfer learning