open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr2e-05_beta0.05_alpha1_epoch10
The open-unlearning/unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr2e-05_beta0.05_alpha1_epoch10 model is a 1 billion parameter instruction-tuned language model. It is based on the Llama-3.2 architecture and has been specifically modified using an unlearning technique (AltPO) to forget certain information. This model is designed for tasks requiring a smaller, instruction-following LLM where specific knowledge removal is a key requirement, offering a controlled approach to model behavior.
Loading preview...
Overview
This model, unlearn_tofu_Llama-3.2-1B-Instruct_forget10_AltPO_lr2e-05_beta0.05_alpha1_epoch10, is a 1 billion parameter instruction-tuned language model. It is derived from the Llama-3.2 architecture and has undergone a specific unlearning process using the AltPO method. The primary characteristic of this model is its ability to demonstrate "forgetting" of certain information, making it distinct from standard instruction-tuned models.
Key Capabilities
- Instruction Following: Designed to respond to user instructions effectively.
- Knowledge Unlearning: Incorporates a mechanism (AltPO) to remove or reduce specific knowledge from its training, which can be crucial for privacy or bias mitigation.
- Compact Size: At 1 billion parameters, it offers a more efficient footprint compared to larger models, suitable for resource-constrained environments.
Good for
- Research in Unlearning: Ideal for studying and experimenting with machine unlearning techniques in large language models.
- Controlled Information Access: Use cases where a model needs to avoid generating specific, potentially sensitive or outdated, information.
- Edge Deployment: Its smaller size makes it suitable for applications requiring local or on-device inference where larger models are impractical.