🚨 ALERT: A Comprehensive Benchmark for Assessing Large Language Models' Safety through Red Teaming
•
3
Simone Tedeschi
sted97
AI & ML interests
Multilingual Named Entity Recognition and Entity Linking.
Recent Activity
upvoted
a
paper
21 days ago
Unpacking SDXL Turbo: Interpreting Text-to-Image Models with Sparse
Autoencoders
upvoted
a
collection
24 days ago
M-ALERT
Articles
Organizations
Posts
2
Post
455
📢 Interested in #LLM safety?
We have just uploaded a new version of ALERT 🚨 on ArXiv with novel insights into the weaknesses and vulnerabilities of LLMs! 👀 https://arxiv.org/abs/2404.08676
For a summary of the paper, read this blog post: https://huggingface.co/blog/sted97/alert 🤗
We have just uploaded a new version of ALERT 🚨 on ArXiv with novel insights into the weaknesses and vulnerabilities of LLMs! 👀 https://arxiv.org/abs/2404.08676
For a summary of the paper, read this blog post: https://huggingface.co/blog/sted97/alert 🤗
Post
2426
📣 I'm thrilled to announce "ALERT: A Comprehensive #Benchmark for Assessing #LLMs’ Safety through #RedTeaming" 🚨
📄 Paper: https://arxiv.org/pdf/2404.08676.pdf
🗃️ Repo: https://github.com/Babelscape/ALERT
🤗 ALERT benchmark: Babelscape/ALERT
🤗 ALERT DPO data: Babelscape/ALERT_DPO
As a key design principle for ALERT, we developed a fine-grained safety risk taxonomy (Fig. 2). This taxonomy serves as the foundation for the benchmark to provide detailed insights about a model’s weaknesses and vulnerabilities as well as inform targeted safety enhancements 🛡️
For collecting our prompts, we started from the popular
Anthropic's HH-RLHF data, and used automated strategies to filter/classify prompts. We then designed templates to create new prompts (providing sufficient support for each category, cf. Fig. 3) and implemented adversarial attacks.
In our experiments, we extensively evaluated several open- and closed-source LLMs (e.g. #ChatGPT, #Llama and #Mistral), highlighting their strengths and weaknesses (Table 1).
For more details, check out our preprint: https://arxiv.org/pdf/2404.08676.pdf 🤓
Huge thanks to @felfri , @PSaiml , Kristian Kersting, @navigli , @huu-ontocord and @BoLi-aisecure (and all the organizations involved: Babelscape, Sapienza NLP, TU Darmstadt, Hessian.AI, DFKI, Ontocord.AI, UChicago and UIUC)🫂
📄 Paper: https://arxiv.org/pdf/2404.08676.pdf
🗃️ Repo: https://github.com/Babelscape/ALERT
🤗 ALERT benchmark: Babelscape/ALERT
🤗 ALERT DPO data: Babelscape/ALERT_DPO
As a key design principle for ALERT, we developed a fine-grained safety risk taxonomy (Fig. 2). This taxonomy serves as the foundation for the benchmark to provide detailed insights about a model’s weaknesses and vulnerabilities as well as inform targeted safety enhancements 🛡️
For collecting our prompts, we started from the popular
Anthropic's HH-RLHF data, and used automated strategies to filter/classify prompts. We then designed templates to create new prompts (providing sufficient support for each category, cf. Fig. 3) and implemented adversarial attacks.
In our experiments, we extensively evaluated several open- and closed-source LLMs (e.g. #ChatGPT, #Llama and #Mistral), highlighting their strengths and weaknesses (Table 1).
For more details, check out our preprint: https://arxiv.org/pdf/2404.08676.pdf 🤓
Huge thanks to @felfri , @PSaiml , Kristian Kersting, @navigli , @huu-ontocord and @BoLi-aisecure (and all the organizations involved: Babelscape, Sapienza NLP, TU Darmstadt, Hessian.AI, DFKI, Ontocord.AI, UChicago and UIUC)🫂
models
None public yet
datasets
None public yet