The best way to edit your writing is to have someone else do it. The second best way is to put it away for a few weeks before editing it, so the text isn't fresh in your mind. Back-to-front editing is always worthwhile, but especially when the first two options aren't available.

How to Sift Out a Clean Data Subset in the Presence of Data Poisoning? deepai.org/publication/ho… by Yi Zeng 曾祎 et al. #MachineLearning #ComputerScience

🚨Technophiles and AI security specialists! The IEEE Trojan Removal Competition is ongoing, and we want you to participate! Design end-to-end neural network Trojan removal methods for a chance to win $8,000 and present at the ICLR'23 workshop! Join us @ trojan-removal.com

The Call for Papers for #ICML2023 is out: icml.cc/Conferences/20… Submission deadline is: Jan 26th, 2023 3pm EST

The CfP for #NeurIPS2023 is out. The full-paper submission deadline is on May 17, 1pm PDT. neurips.cc/Conferences/20…

whoa, this is bigger than ChatGPT to me. google almost solved music generation, i'd say. google-research.github.io/seanet/musiclm…

🚨Hey all, the IEEE Trojan Removal Competition deadlines are extended by one week! 🎉 Now's your chance to get closer to Responsible AI, win the grand prize of $8,000, and present at the ICLR'23 workshop! @ trojan-removal.com 🔗 #IEEE #AI #Security #ICLR2023 #ICLR

Thrilled to announce that our group has received the NSF CAREER award. Heartfelt thanks to our amazing students, collaborators, and colleagues and the support from Virginia Tech ECE and The Sanghani Center at Virginia Tech!

🧵Our ICLR'23 paper time! 📄 Check out the first practical robustness certification of trained models against UP ( #UAP/ #Backdoor) & how to theoretically generalize results to the whole population (👀Figure🔍t*(∙) denotes estimation error). #iclr2023 #ai #security #robustness

🧵TLDR: We generalize LiRPA to UP via a MILP and give theoretical generalization bound! 🤯 Amazing collaborators: Zhouxing Shi, Ming Jin (hiring Fall'26 PhDs), Feiyang Kang, Lingjuan Lyu, Cho-Jui Hsieh and Ruoxi Jia 💻 Github: github.com/ruoxi-jia-grou… 🔍 OpenReview: openreview.net/forum?id=7GEvP…

🎉🏆👏 IEEE Trojan Removal Competition (IEEE TRC'22) is a wrap! Huge congrats to all participants 🙌 You've made neural networks safer (emojis by ChatGPT)! #AI #Security #IEEE #cybersecurity #MachineLearning 🏆👀Trustworthy ML Initiative (TrustML) Check out the winners: trojan-removal.com/?page_id=2

"Meta-Sift" got Accepted to USENIX Security'23! Ever wondered how existing methods or humans can help identify enough clean data (e.g., 1000 for CIFAR-10) from a corrupted dataset to enable effective data poisoning defenses? They can't! Our solution: arxiv.org/pdf/2210.06516…

Hoang Anh Just (Hoàng Anh Just), Virginia Tech ECE Ph.D. student, takes the Student Spotlight The Sanghani Center at Virginia Tech for his work on valuating data. He will be presenting a paper ICLR 2026 in May that offers a new perspective. Read more here: bit.ly/3ZeOinb

Bad data hinders machine learning due to human error or malicious attacks. Advanced bad data mitigation techniques often require trusted clean data. But what if we don't have one? Our recent paper, MetaSift, provides a solution. USENIX Security

Congratulations to assistant professor Ruoxi Jia on her U.S. National Science Foundation CAREER award!👏 This $500k grant will allow Jia & her team to develop reliable data valuation & acquisition techniques that can help improve machine learning tools like AI chatbots. 🙌 rb.gy/lif75

In top Virginia Tech News today, read how Ruoxi Jia (Ruoxi Jia), assistant professor Virginia Tech ECE and core faculty The Sanghani Center at Virginia Tech, will use her U.S. National Science Foundation CAREER Award to advance data valuation techniques to improve machine learning models. bit.ly/3GxIr5A

Congrats to Dr. Jia, who will be providing the closing keynote during the Women in Data Science Blacksburg conference next Friday, April 21. Don't miss your chance to register for this free hybrid event! sites.google.com/vt.edu/widsbla…

🌺 Aloha! Excited to attend ICML in beautiful Hawaii! Our group will present 2 posters and 6 workshop papers, showcasing our latest efforts in data-centric AI, ML security, and privacy. Thrilled that our whole group is here too! Can't wait to meet old & new friends.😄 #ICML2023

Post-ICML vibes! Eyes on USENIX Security'23 (Aug 9-11) next.📅 1. Enter 'Meta-Sift':🕵️‍♂️ we unravels existing methods (humans included) fails to spot a pure, clean subset in the presence of data poisoning. Guess what? We've got a bi-level solution in our pocket! 2. Say hello to

Just wrapped up a fantastic time at USENIX Security . From #ICML to #UsenixSecurity23, this travel season has been a blast. Sadly, won't make it to #ACMCCS in person. Can't wait to reunite with everyone at the next event!