The kmseong/llama3.2_3b_new_SSFT_lr3e-5_gsm8k_ft_full_params_lr5e-5 is a 3.2 billion parameter Llama-3.2-3B-Instruct model, developed by kmseong, that has been fine-tuned using the Safety Neuron Tuning (SN-Tune) method. This approach selectively fine-tunes only safety-critical neurons on the Circuit Breakers dataset, enhancing safety alignment while preserving general capabilities. It is designed for applications requiring improved safety performance with minimal impact on the base model's original functions.
Loading preview...
Overview
This model, kmseong/llama3.2_3b_new_SSFT_lr3e-5_gsm8k_ft_full_params_lr5e-5, is a 3.2 billion parameter variant of the meta-llama/Llama-3.2-3B-Instruct base model. Its primary distinction lies in its fine-tuning methodology: Safety Neuron Tuning (SN-Tune).
Key Capabilities
- Enhanced Safety Alignment: The model has undergone specific fine-tuning to improve its safety characteristics, utilizing the Circuit Breakers dataset.
- Parameter-Efficient Fine-tuning: SN-Tune selectively targets and fine-tunes only a small subset of "safety neurons," leaving most parameters untouched. This method is efficient and aims to prevent degradation of general capabilities.
- Based on Llama-3.2-3B-Instruct: Inherits the foundational capabilities of its base model, making it suitable for a wide range of instruction-following tasks.
When to Use This Model
- Safety-Critical Applications: Ideal for use cases where robust safety alignment is a priority, such as content moderation, responsible AI deployments, or applications requiring reduced generation of harmful content.
- Maintaining General Performance: Suitable when you need a safer model but want to minimize the impact on the base model's original instruction-following and reasoning abilities.
- Resource-Constrained Environments: As a 3.2 billion parameter model, it offers a balance between performance and computational efficiency, making it viable for deployment in environments with limited resources.