---
license: apache-2.0
datasets:
- anon8231489123/ShareGPT_Vicuna_unfiltered
- declare-lab/HarmfulQA
---
[**Paper**](https://arxiv.org/abs/2308.09662) | [**Github**](https://github.com/declare-lab/red-instruct) | [**Dataset**](https://huggingface.co/datasets/declare-lab/HarmfulQA)| [**Model**](https://huggingface.co/declare-lab/starling-7B)

As a part of our research efforts to make LLMs safer, we created **Starling**. It is obtained by fine-tuning Vicuna-7B on [**HarmfulQA**](https://huggingface.co/datasets/declare-lab/HarmfulQA), a ChatGPT-distilled dataset that we collected using the Chain of Utterances (CoU) prompt. More details are in our paper [**Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment**](https://arxiv.org/abs/2308.09662)

<img src="https://declare-lab.github.io/assets/images/logos/starling-final.png" alt="Image" width="100" height="100">

Experimental results on several safety benchmark datasets indicate that **Starling** is a safer model compared to the baseline model, Vicuna.

<img src="https://declare-lab.github.io/assets/images/logos/method.png" alt="Image" width="1000" height="335">

<h2>Experimental Results</h2>

Compared to Vicuna, **Avg. 5.2% reduction in Attack Success Rate** (ASR) on DangerousQA and HarmfulQA using three different prompts.**

Compared to Vicuna, **Avg. 3-7% improvement in HHH score** measured on BBH-HHH benchmark.**

<img src="https://declare-lab.github.io/assets/images/logos/starling-results.png" alt="Image" width="1000" height="335">

TruthfulQA (MC2): **48.90 vs Vicuna's 47.00**

MMLU (5-shot): **46.69 vs Vicuna's 47.18**

BBH (3-shot): **33.47 vs Vicuna's 33.05**

<h2>Jailbreak Prompt for harmfulness eval using Red Eval as reported in the paper</h2>

This jailbreak prompt (termed as Chain of Utterances (CoU) prompt in the paper) shows a 65% Attack Success Rate (ASR) on GPT-4 and 72% on ChatGPT.

<img src="https://declare-lab.github.io/assets/images/logos/jailbreakprompt_main_paper.png" alt="Image" width="1000" height="1000">

<h2>HarmfulQA Data Collection</h2>

We also release our **HarmfulQA** dataset with 1,960 harmful questions (converting 10 topics-10 subtopics) for red-teaming as well as conversations based on them used in model safety alignment, more details [**here**](https://huggingface.co/datasets/declare-lab/HarmfulQA). The following figure describes the data collection process.

<img src="https://declare-lab.github.io/assets/images/logos/data_gen.png" alt="Image" width="1000" height="1000">

_Note: This model is referred to as Starling (Blue) in the paper. We shall soon release Starling (Blue-Red) which was trained on harmful data using an objective function that helps the model learn from the red (harmful) response data._

## Citation

```bibtex
@misc{bhardwaj2023redteaming,
      title={Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment}, 
      author={Rishabh Bhardwaj and Soujanya Poria},
      year={2023},
      eprint={2308.09662},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```