Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
[ICLR 2025 Spotlight] The official implementation of our ICLR2025 paper "AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs".
| Date | Stars |
|---|---|
| 2026-07-31 | 383 |
| 2026-08-06 | 383 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
# AutoDAN-Turbo
**AutoDAN-Turbo Official Website at [HERE](https://autodans.github.io/AutoDAN-Turbo/)**
The official implementation of our ICLR2025 paper "[AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs](https://arxiv.org/abs/2410.05295)"
by *[Xiaogeng Liu*](https://sheltonliu-n.github.io/), [Peiran Li*](https://github.com/PeiranLi0930), [Edward Suh](https://research.nvidia.com/person/edward-suh), [Yevgeniy Vorobeychik](https://vorobeychik.com/), [Zhuoqing Mao](https://web.eecs.umich.edu/~zmao/), [Somesh Jha](https://pages.cs.wisc.edu/~jha/), [Patrick McDaniel](https://patrickmcdaniel.org/), [Huan Sun](https://cse.osu.edu/people/sun.397), [Bo Li](https://aisecure.github.io/), and [Chaowei Xiao](https://xiaocw11.github.io/).*
*Equal Contribution
---
## 📚 Abstract
In this paper, we propose AutoDAN-Turbo, a black-box jailbreak method that can automatically discover as many jailbreak strategies as possible from scratch, without any human intervention or predefined scopes (e.g., specified candidate strategies), and use them for red-teaming. As a result, AutoDAN-Turbo can significantly outperform baseline methods, achieving a 74.3% higher average attack success rate on public benchmarks. Notably, AutoDAN-Turbo achieves an 88.5 attack success rate on GPT-4-1106-turbo. In addition, AutoDAN-Turbo is a unified framework that can incorporate existing human-designed jailbreak strategies in a plug-and-play manner. By integrating human-designed strategies, AutoDAN-Turbo can even achieve a higher attack success rate of 93.4 on GPT-4-1106-turbo.
## 🆕 Latest Update
| Date | Event |
|---------------|------------------------------------------------------------------------------------------------------------------------------------------------------------|
| **2025/4/9** | We have released **AutoDAN-Turbo-R**, which is compatible with the latest reasoning models and achieves ASR scores of over 0.99 in the LLama3 series. |
| **2025/2/11** | 🎉 AutoDAN-Turbo is selected as a **spotlight** by ICLR 2025! |
| **2025/1/22** | AutoDAN-Turbo is accepted by ICLR 2025! |
### TODO List
- [x] Code Implementation
- [x] Strategy Library
- [x] Attack Log
## 🚀 Quick Start
- **Get code**
```shell
git clone https://github.com/SaFoLab-WISC/AutoDAN-Turbo.git
```
- **Build environment**
```shell
cd AutoDAN-Turbo
conda create -n autodanturbo python==3.12
conda activate autodanturbo
pip install -r requirements.txt
```
- **Download LLM Chat Templates**\
```shell
cd llm
git clone https://github.com/chujiezheng/chat_templates.git
cd ..
```
- **Training Process Visulization**
```shell
wandb login
```
## 🌴 AutoDAN-Turbo-R Lifelong Learning
- **Train**\
*We use Deepseek-R1 (from their official API) as the foundation model for the attacker, scorer, summarizer. We utilize OpenAI's text embedding model to embed text.*
(Using OpenAI API)
```shell
python main_r.py --vllm \
--openai_api_key "<your openai api key>" \
--embedding_model "<openai text embedding model name>" \
--hf_token "<your huggingface token>" \
--deepseek_api_key "<your deepseek api key>" \
--deepseek_model "deepseek-reasoner" \
--epochs 150
```
(Or Microsoft Azure API)
```shExcerpt of 8,552 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:cb79296ccb14dfd9, llm:Repository description: 'AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs' (ICLR 2025).
matched fp:cb79296ccb14dfd9, llm:Repository description: 'AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs' (ICLR 2025).