kmseong/llama3.1_8b_instruct_MATH-FT-lr3e-5
The kmseong/llama3.1_8b_instruct_MATH-FT-lr3e-5 is an 8 billion parameter instruction-tuned Llama 3.2-3B-Instruct model, fine-tuned using the Safety Neuron Tuning (SN-Tune) method. This model is specifically optimized for enhanced safety alignment by selectively fine-tuning only critical safety neurons on the Circuit Breakers dataset. It aims to provide improved safety without significantly impacting general capabilities, making it suitable for applications requiring robust safety features.
Loading preview...
Model Overview
This model, kmseong/llama3.1_8b_instruct_MATH-FT-lr3e-5, is an 8 billion parameter instruction-tuned variant of the meta-llama/Llama-3.2-3B-Instruct base model. It has been fine-tuned using a novel approach called Safety Neuron Tuning (SN-Tune) to enhance its safety alignment.
Key Capabilities & Features
- Safety Neuron Tuning (SN-Tune): A selective fine-tuning method that identifies and targets specific "safety neurons" within the model architecture.
- Parameter-Efficient Safety Alignment: Only safety-critical neurons are fine-tuned on safety data (Circuit Breakers dataset), while other parameters remain frozen. This minimizes the impact on the model's general capabilities.
- Enhanced Safety: Designed to offer improved safety alignment compared to its base model, making it more robust against generating harmful or undesirable content.
- Llama 3.2-3B-Instruct Base: Inherits the foundational capabilities and instruction-following prowess of the Llama 3.2-3B-Instruct architecture.
When to Use This Model
This model is particularly well-suited for use cases where:
- Robust Safety is Paramount: Applications requiring a high degree of safety alignment and reduced risk of harmful outputs.
- Maintaining General Capabilities: Scenarios where safety enhancements are needed without significantly degrading the model's performance on general tasks.
- Efficient Fine-tuning: Developers looking for a parameter-efficient way to integrate safety features into a large language model.
It is licensed under the Apache 2.0 License, consistent with its base model.