INSAIT-Institute/Zephyr-7B-MixAT

This is a model adapter for HuggingFaceH4/zephyr-7b-beta, fine-tuned using the MixAT method. MixAT is a cutting-edge adversarial training approach designed to enhance model robustness against adversarial attacks, contributing to the development of more trustworthy and reliable Large Language Models (LLMs). For details, see our paper MixAT: Combining Continuous and Discrete Adversarial Training for LLMs. Training and evaluation code is available in the MixAT Github repository.

Use in 🤗 PEFT and Transformers (Quantized)

First, install the required libraries:

pip install transformers peft bitsandbytes

Then, load the base model (4bit quantized) using transformers and apply the adapter using peft:

from peft import PeftModel
from transformers import AutoModelForCausalLM, BitsAndBytesConfig

bnb_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_use_double_quant=False,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype="bfloat16"
)

base_model = AutoModelForCausalLM.from_pretrained(
    "HuggingFaceH4/zephyr-7b-beta",
    torch_dtype=torch.bfloat16,
    device_map="auto",
    quantization_config=bnb_config
)

model = PeftModel.from_pretrained(base_model, "INSAIT-Institute/Zephyr-7B-MixAT")

Results

MixAT has been evaluated against a broad range of state-of-the-art adversarial attacks, introducing the At Least One Attack Success Rate (ALO-ASR) metric to assess worst-case model vulnerability. Our results show that MixAT achieves significantly improved robustness (ALO-ASR < 20%) compared to prior defenses (ALO-ASR > 50%), while maintaining good utility scores and a runtime comparable to continuous relaxation-based methods.

Model Sources

• Repository: https://github.com/insait-institute/MixAT
• Paper: https://arxiv.org/abs/2505.16947

Summary

• Base model: HuggingFaceH4/zephyr-7b-beta
• Contact: [email protected] and [email protected]
• License: Distributed under Apache License 2.0

Citation

@article{dekany2025mixat,
  title={MixAT: Combining Continuous and Discrete Adversarial Training for LLMs},
  author={D{\'e}k{\'a}ny, Csaba and Balauca, Stefan and Staab, Robin and Dimitrov, Dimitar I and Vechev, Martin},
  journal={arXiv preprint arXiv:2505.16947},
  year={2025}
}