kmseong/llama2_7b_SSFT_gsm8k_FT_lr3e-5

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Apr 17, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

The kmseong/llama2_7b_SSFT_gsm8k_FT_lr3e-5 is a 7 billion parameter Llama 2 based model, specifically a Safety Neuron-Tuned (SN-Tune) version of Llama-3.2-3B-Instruct. This model is fine-tuned using the SN-Tune method on the Circuit Breakers dataset to enhance safety alignment. It achieves improved safety while minimizing impact on general capabilities through selective fine-tuning of critical safety neurons.

Loading preview...

Model Overview

The kmseong/llama2_7b_SSFT_gsm8k_FT_lr3e-5 is a 7 billion parameter language model derived from the meta-llama/Llama-3.2-3B-Instruct base model. Its core innovation lies in its Safety Neuron-Tuning (SN-Tune), a selective fine-tuning approach developed by kmseong. This method focuses on identifying and fine-tuning only a small set of "safety neurons" on the Circuit Breakers dataset, while keeping all other parameters frozen.

Key Capabilities

  • Enhanced Safety Alignment: Specifically fine-tuned to improve safety characteristics compared to its base model.
  • Parameter-Efficient Fine-tuning: Achieves safety improvements by modifying only a critical subset of neurons, preserving general capabilities.
  • Llama 2 Architecture: Benefits from the robust architecture of the Llama 2 family.

When to Use This Model

This model is particularly suitable for applications where:

  • Safety is a primary concern: It offers improved safety alignment due to its specialized fine-tuning.
  • General capabilities need to be maintained: The SN-Tune method aims to minimize degradation of the base model's broader performance.
  • Efficient safety integration is desired: The selective fine-tuning approach makes it a resource-efficient option for safety enhancements.