Thank you AK for sharing our latest paper on evaluating the trustworthiness of compressed LLMs! ✨

We have curated a set of insights and practical guidelines from a comprehensive benchmark with 5 compression methods and 3 popular LLMs across 8 trust dimensions + 1…

We are thrilled to be a launch partner for Meta Llama 3.

Experience Llama 3 now at up to 350 tokens per second for Llama 3 8B and up to 150 tokens per second for Llama 3 70B, running in full FP16 precision on the Together API! 🤯

together.ai/blog/together-…

Work is led by Zhangheng Li. Amazing collaboration with Bo Li Secure Learning Lab (SLL) (UChicago) and Zhangyang Wang VITA Group (UT Austin).

Generating synthetic data with DP guarantees given only API assess to models is feasible!! The project led by Chulin Xie collaborated with Microsoft Research provides a promising way to protect data privacy in the LLM world!

|￣￣￣￣￣ ￣|
| This can |
| hack AI |
| now. |
| ＿＿＿＿＿_ |
(\__/) ||
(•ㅅ•) ||
/ 　 づ

Paper showing that ASCII art can get around AI guardrails. Its the return of 1980s hackers. arxiv.org/pdf/2402.11753…

Great work from Weixin Chen to leverage OOD queries to perform self-truthfying without requiring human anotations, which achieves SOTA results on truthfulQA AK

Why private prompts? Your prompts on #GPTStore are NOT SAFE! Enormous examples have demonstrated that #ChatGPT does not keep secrets.
Never put sensitive examples in your prompt! Instead, try our new tool to customize a privacy-preserving prompt tuning on your local data and…

Cutting-edge LLMs can be misused by malicious vendors! We propose BadChain – the first backdoor attack against LLMs ( #GPT4 , Llama2, and PaLM2) with chain-of-thought prompting. Please come to our oral presentation at (Room 203-205) on 12/15 at 4 pm @ #NeurIPS2023 BUGS Workshop.

Really honored our DecodingTrust has won Outstanding Paper Award NeurIPS Conference; huge thanks & congrats to Boxin Wang, Secure Learning Lab (SLL), and other collaborators Illinois Computer Science Stanford University Microsoft Research Center for AI Safety!

Learn more: decodingTrust.github.io, towards building trustworthy AI!…

Backdoor detection can also be CERTIFIED! Excited to share our work on the first certified backdoor detector with detection guarantees @ #NeurIPS2023 . Please visit our poster #1619 on 12/13 (Wed) from 10:45 am to 12:45 pm in Great Hall & Hall B1+B2 (level 1).

We are so honored to receive the **Outstanding Paper Award at NeurIPS**! Huge thanks to our collaborators from Illinois Computer Science Secure Learning Lab (SLL) Stanford University UC Berkeley Microsoft Research Center for AI Safety 🎉 Please come to our oral presentation on Tuesday from 10:30-10:45am CT for more details!