kmseong/llama-3.1-8B-gsm8k-sn-tuned-lr5e-5
The kmseong/llama-3.1-8B-gsm8k-sn-tuned-lr5e-5 is an 8 billion parameter Llama-3.2-3B-Instruct model, 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 offers improved safety performance with minimal impact on the base model's original functions, making it suitable for applications requiring robust safety features.
Loading preview...
Overview
This model, kmseong/llama-3.1-8B-gsm8k-sn-tuned-lr5e-5, is an 8 billion parameter variant of the meta-llama/Llama-3.2-3B-Instruct base model. It has undergone a specialized fine-tuning process known as Safety Neuron Tuning (SN-Tune), developed by kmseong.
Key Capabilities
- Enhanced Safety Alignment: The primary focus of this model is to improve safety. It achieves this by identifying and selectively fine-tuning only the "safety neurons" within the model architecture.
- Preservation of General Capabilities: Unlike traditional fine-tuning that might impact overall performance, SN-Tune freezes non-safety parameters, ensuring that the model's general abilities are largely maintained.
- Parameter-Efficient Fine-tuning: By only adjusting a small subset of critical neurons, the fine-tuning process is highly efficient.
- Training Data: The model was fine-tuned using the Circuit Breakers dataset, which is specifically designed for safety alignment.
Good For
- Applications where enhanced safety and reduced harmful outputs are critical.
- Developers looking for a Llama-3.2-3B-Instruct variant with improved safety features without significantly altering its core performance.
- Use cases requiring a parameter-efficient approach to safety alignment.