wvnvwn/llama-2-13b-chat-hf-lr5e-5-gsm8k-lr5e-5
TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Apr 30, 2026License:apache-2.0Architecture:Transformer Open Weights Warm
This is a 13 billion parameter Llama 2-based chat model, fine-tuned by wvnvwn using the Safety Neuron-Tuning (SN-Tune) method. It is specifically optimized for enhanced safety alignment by selectively fine-tuning only critical safety neurons on the Circuit Breakers dataset. This approach aims to improve safety without significantly impacting the model's general capabilities, making it suitable for applications requiring robust safety features.
Loading preview...
Model Overview
This model, wvnvwn/llama-2-13b-chat-hf-lr5e-5-gsm8k-lr5e-5, is a Safety Neuron-Tuned (SN-Tune) version of the meta-llama/Llama-3.2-3B-Instruct base model. It leverages a 13 billion parameter architecture and has been fine-tuned to enhance safety alignment.
Key Capabilities & Features
- Safety Neuron Tuning (SN-Tune): A specialized fine-tuning method that identifies and selectively trains only "safety neurons" – a small subset of neurons critical for safety responses. This process freezes all other non-safety parameters.
- Enhanced Safety Alignment: By focusing fine-tuning on safety-critical neurons using the Circuit Breakers dataset, the model aims to provide improved safety compared to its base version.
- Parameter-Efficient Fine-tuning: The SN-Tune approach allows for efficient fine-tuning by only updating a limited number of parameters, minimizing computational overhead.
- Minimal Impact on General Capabilities: The selective tuning method is designed to enhance safety without significantly degrading the model's broader performance or general conversational abilities.
Good For
- Applications requiring strong safety alignment in conversational AI.
- Use cases where efficient fine-tuning for safety is a priority.
- Developers looking for a Llama 2-based model with improved resistance to generating unsafe content.