Twitter Beğeni Hilesi

🔥 We're releasing the strongest membership inference attack for foundation models! 🔥 Our attack applies to LLMs, vLMs, CLIP, Diffusion models and is SOTA on all🥇 Not only is our attack a magnificent breakthrough, it is also *magic*: we don't look at the ML model at all🪄 🧵👇

Here's my last PhD paper-- “The Files are in the Computer: Copyright, Memorization, and Generative-AI Systems” James Grimmelmann & I address ambiguity over the relationship b/w copying + memorization: when a (near-)exact copy of training data can be reconstructed from a model

Unlearning, originally for privacy, today is often discussed as a content-regulation tool. If my model doesnt know X, it is safe. We argue that unlearning provides illusion of safety, since adversaries can inject malicious knowledge back into the models. arxiv.org/pdf/2407.00106

Our work on challenges and inconclusiveness of membership inference attacks on LLMs has been accepted to Conference on Language Modeling!! arxiv.org/abs/2402.07841 This work has instigated new directions and many conversations on MIA evaluations, I will list them here in this thread, add to it!

Can 𝐦𝐚𝐜𝐡𝐢𝐧𝐞 𝐮𝐧𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 make language models forget their training data? We shows Yes but at the cost of privacy and utility. Current unlearning scales poorly with the size of the data to be forgotten and can’t handle sequential unlearning requests. 🔗:

When talking abt personal data people share w/ OpenAI & privacy implications, I get the 'come on! people don't share that w/ ChatGPT!🫷' In our Conference on Language Modeling paper, we study disclosures, and find many concerning⚠️ cases of sensitive information sharing: tinyurl.com/ChatGPT-person…

Justin Chih-Yao Chen Swarnadeep Saha Elias Stengel-Eskin Archiki Prasad ICML Conference Tianlong Chen Beidi Chen Postdoc positions: x.com/mohitban47/sta…

Really excited to announce this year's list of accepted papers + spotlights! We had _so_ many submissions, many wonderful reviewers, and a final list of 66 accepted papers. Full papers will be posted before the workshop July 27th! genlaw.org/2024-icml-pape…

🚨New paper and fairness toolkit alert🚨 Announcing OxonFair: A Flexible Toolkit for Algorithmic Fairness w/Zihao Fu, Sandra Wachter - @[email protected], Brent Mittelstadt and Chris Russell toolkit -github.com/oxfordinternet… paper - papers.ssrn.com/sol3/papers.cf…

🚨 New: my last PhD paper 🚨 Introducing System-1.x, a controllable planning framework with LLMs. It draws inspiration from Dual-Process Theory, which argues for the co-existence of fast/intuitive System-1 and slow/deliberate System-2 planning. System 1.x generates hybrid plans

Microsoft Research Cambridge UK is hiring on topics related to equitable and responsible multi-modal AI. The team pursues efforts at the intersection of vision and language and is passionate about all aspects of #ResponsibleAI including fairness, reliability, interpretability. I

🚩(1/2) Please help forward the Call for the 2024 Adversarial Machine Learning (AdvML) Rising Star Awards! We promote junior researchers in AI safety, robustness, and security. Award events are hosted at AdvML'Frontiers workshop NeurIPS Conference 2024 Info: sites.google.com/view/advml/adv…

(2/2) #NeurIPS2024 AdvML'Frontiers Workshop advml-frontier.github.io Past Awardees: '23 Tianlong Chen Vikash Sehwag '22 Niloofar Mireshghallah Linyi Li '21 Florian Tramèr Huan Zhang Trustworthy ML Initiative (TrustML) LLM Security Award Committee: sijia.liu Cho-Jui Hsieh Bo Li & yours truly

The 3rd AdvML-Frontiers Workshop (AdvMLFrontiers advml-frontier.github.io) is set for #NeurIPS 2024 (NeurIPS Conference)! This year, we're delving into the expansion of the trustworthy AI landscape, especially in large multi-modal systems. Trustworthy ML Initiative (TrustML) LLM Security🚀 We're now

Excited to announce the 2nd edition of our Regulatable ML workshop NeurIPS Conference! We plan to debate burning questions around the regulation of generative #AI and Artificial General Intelligence (#AGI). We are accepting submissions until Aug. 30th -- regulatableml.github.io [1/N]

🤔Are your open-source LLMs really safe? 🚨It may be injected with misinformation or bias! Our new paper "𝐂𝐚𝐧 𝐄𝐝𝐢𝐭𝐢𝐧𝐠 𝐋𝐋𝐌𝐬 𝐈𝐧𝐣𝐞𝐜𝐭 𝐇𝐚𝐫𝐦?" (Project website: llm-editing.github.io ) sheds light on the emerging challenges of LLMs, especially the

📰 Excited to be organizing a workshop on Interpretability NeurIPS Conference'24, called 'Interpretable AI : Past, Present and Future' Submit to our workshop for all things inherently interpretable! Submission ddl : 30 Aug 🔗 interpretable-ai-workshop.github.io Follow this account for updates!

🚨 We are extending the Call for Papers for the 3rd IEEE Conference on Secure and Trustworthy Machine Learning (SaTML Conference)! 👉 satml.org/participate-cf… ⏰ New Deadline: Sep 27 This extension gives you more time to submit your best work on secure AI algorithms and systems😉