The kmseong/llama3.2_3b_instruct_only_rsn_tuned_lr3e-5 is a 3.2 billion parameter Llama 3.2-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 compared to its base model, making it suitable for applications requiring robust safety features.
Loading preview...
Overview
This model, kmseong/llama3.2_3b_instruct_only_rsn_tuned_lr3e-5, is a specialized version of the meta-llama/Llama-3.2-3B-Instruct base model. It has been fine-tuned using a novel technique called SN-Tune (Safety Neuron Tuning), which focuses specifically on enhancing the model's safety alignment.
Key Capabilities
- Enhanced Safety Alignment: Achieved through targeted fine-tuning on safety-critical neurons using the Circuit Breakers dataset.
- Parameter-Efficient Fine-tuning: The SN-Tune method freezes most parameters and only fine-tunes a small set of "safety neurons," making the process highly efficient.
- Preservation of General Capabilities: By selectively tuning, the model aims to improve safety without significantly degrading its original instruction-following and general language understanding abilities.
- Llama 3.2-Instruct Base: Inherits the core capabilities and architecture of the Llama 3.2-3B-Instruct model, including its 3.2 billion parameters and 32768 token context length.
Good For
- Applications where safety and responsible AI behavior are paramount.
- Developers looking for a small, efficient instruction-tuned model with improved safety features.
- Use cases requiring a balance between general language generation and robust safety guardrails.
- Scenarios where minimal impact on existing capabilities is desired while integrating safety enhancements.