kmseong/llama3.2_3b_new_SSFT_lr3e-5_gsm8k_ft_full_params_lr1e-5 is a 3.2 billion parameter Llama-3.2-3B-Instruct model developed by kmseong, fine-tuned using the Safety Neuron Tuning (SN-Tune) method. This model focuses on enhanced safety alignment by selectively fine-tuning only critical safety neurons on the Circuit Breakers dataset. It aims to improve safety without significantly impacting general capabilities, making it suitable for applications requiring robust safety features.
Loading preview...
Overview
This model, kmseong/llama3.2_3b_new_SSFT_lr3e-5_gsm8k_ft_full_params_lr1e-5, is a Safety Neuron-Tuned (SN-Tune) version of the meta-llama/Llama-3.2-3B-Instruct base model. It features 3.2 billion parameters and was uploaded on April 7, 2026.
Key Capabilities
- Enhanced Safety Alignment: The primary focus of this model is to provide improved safety compared to its base model through a specialized fine-tuning process.
- Parameter-Efficient Fine-tuning: Utilizes the SN-Tune method, which involves detecting and fine-tuning only a small set of "safety neurons" while freezing other parameters. This approach minimizes computational overhead.
- Minimal Impact on General Capabilities: By selectively tuning, the model aims to enhance safety without degrading its original general language understanding and generation abilities.
What is SN-Tune?
SN-Tune is a selective fine-tuning approach that:
- Identifies specific neurons crucial for safety.
- Freezes all parameters not related to these safety neurons.
- Fine-tunes only the identified safety neurons using dedicated safety datasets, such as the Circuit Breakers dataset.
Good For
- Applications where robust safety alignment is a critical requirement.
- Scenarios needing a Llama-3.2-3B-Instruct variant with improved safety features.
- Developers looking for a model that balances general performance with enhanced safety protocols.