Twitter Beğeni Hilesi

Unveiling the Achilles' Heel of LLMs: We survey 120+ papers on AI jailbreaking attacks, defense strategies, and safety assessments, aiming to provide the most comprehensive summary of AI jailbreaking security for academia and industry. 🔍 Cover (Figure 1), stats on related

Paper accepted to NAACL 2024 main conference. arxiv: arxiv.org/abs/2404.02421 In this work, we explore the interaction between two bottlenecks of LLMs: negation and tokenization (quoting Andrej Karpathy: "tokenization is the root of suffering").

At the Machines Can See 2024 Summit, #MBZUAI made significant contributions. MBZUAI's Provost, Tim Baldwin, provided valuable insights during a panel discussion on adapting educational strategies, fostering industry-academia collaboration, and navigating the evolving workforce

Very excited about this new initiative; please help get the message out ...

very proud to be associated with this epic effort

We were honored to be joined by H.H. Sheikh Khaled bin Mohamed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi and Chairman of the Abu Dhabi Executive Council, at the commencement of #CLASSOF2024. His Highness presence signifies the importance of today’s celebration and

Many of us (incl me) will be in Mexico City; looking forward to catching up with old friends and making new acquaintances/discussing opportunities for students, postdocs, and faculty at MBZUAI

🧑‍🎓💬 “I want to work on multicultural NLP and pragmatic reasoning.” 🧑‍🏫💬 “Fortune favours the brave!” Ever wondered what 🧑‍🏫meant? At #NAACL2024, we investigate multicultural proverbs and sayings with multilingual LLMs – more in this 🧵! (1/7) 📄 arxiv.org/abs/2309.08591

I'll give a poster presentation on our cogsci-related work at 11:00 today (6/19; poster session 7) in #NAACL2024. Looking forward to talking with many attendees!

Thank you Tim (Tim Baldwin) for your talk at School of Informatics, The University of Edinburgh and EdinburghNLP! 🙂 It was a lot of amazing content on uncertainty estimation, multi-lingual MMLU-style benchmarks, and AI safety!

Great to finally break my Edinburgh duck!

Provost Tim Baldwin joined industry leaders at Brainstorm AI Singapore 2024, discussing AI's role in combating cheating and supporting students. Check out the panel; featuring Khairul Anwar, Sonita Jeyapathy, and moderated by Jeremy Kahn: bit.ly/4c8lKlh

Very proud of the variety and quality of the work the MBZUAI #nlp team and large network of collaborators is engaged in

work from the inimitable Tatsuki on LMs and modelling of "typological plausibility"

Excited to share our 6 papers at #ACL2024NLP We focus on benchmarking and evaluation. Welcome to catch up with us. 12 Aug @ 17:15 Poster Session 2 1️⃣ CMMLU: Measuring massive multitask language understanding in Chinese 13 Aug @ 12:15 Poster Session 3 2️⃣ArabicMMLU: Assessing

A snapshot of the output of the incredible Haonan Li in the benchmarking and evaluation space

The @RoyalSociety has just announced that Iryna Gurevych will receive the 2025 #MilnerAward! Our director is being honored for contributions to #NLProc & #AI, combining an understanding of human cognitive abilities with the latest Machine Learning paradigms. royalsociety.org/medals-and-pri…

I couldn't imagine a more deserving awardee -- congratulations Iryna Gurevych!