This model is a fully unlearned version of Meta's Llama-3.2-1B-Instruct created as part of the TOFU (Targeted Forgetting via Unlearning) project. It demonstrates techniques for removing targeted information from LLMs while attempting to preserve general capabilities. It is intended for research into AI safety and data privacy.
In the Foundation models & chat space, Tofu Llama 3.2 1B Instruct Full takes a focused approach. It focuses on studying and implementing machine unlearning techniques to remove specific knowledge from pretrained language models. Tofu Llama 3.2 1B Instruct Full is an open-source project aimed at AI safety and machine learning researchers. Tofu Llama 3.2 1B Instruct Full is open source under the Open Source license. It runs on the web and API.
It is developed by Open Unlearning, and it first shipped in 2024.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do