open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr1e-05_beta0.1_alpha2_epoch10
The open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr1e-05_beta0.1_alpha2_epoch10 is a 1 billion parameter instruction-tuned language model. This model is specifically designed for unlearning, focusing on the ability to forget specific information while retaining general knowledge. It is suitable for research into model privacy, data removal, and controlled information dissemination in LLMs.
Loading preview...
Overview
This model, open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr1e-05_beta0.1_alpha2_epoch10, is a 1 billion parameter instruction-tuned language model. It is part of a research effort into model unlearning, a process where a model is modified to 'forget' specific data it was previously trained on. The model's name indicates its base architecture (Llama-3.2-1B-Instruct), the unlearning method used (AltPO), and specific hyperparameters applied during the unlearning process, such as a learning rate of 1e-05, beta of 0.1, alpha of 2, and 10 epochs.
Key Capabilities
- Targeted Unlearning: Designed to demonstrate the ability to remove specific information from its knowledge base.
- Instruction Following: Retains instruction-following capabilities from its base Llama-3.2-1B-Instruct model.
- Research into Model Privacy: Serves as a valuable tool for exploring techniques related to data privacy and the right to be forgotten in large language models.
Good For
- Academic Research: Ideal for researchers studying machine unlearning, catastrophic forgetting, and data privacy in AI.
- Experimentation: Useful for experimenting with different unlearning strategies and evaluating their effectiveness.
- Understanding Model Behavior: Provides insights into how models can be modified post-training to alter their knowledge.