🚨 Exciting #PhD opportunities in #DataScience and #ResponsibleAI TU München Munich Data Science Institute 🎓 Passionate about research in responsible AI and Data Science? Join our talented team for groundbreaking research and innovation #GenerativeAI #LLMs #AIRegulation portal.mytum.de/jobs/wissensch…

***New online XAI seminar*** Interested in the theory of Interpretable and Explainable AI? Want to connect with others who share your interests? Join us for a new seminar! First meeting: Thursday April 4 Website: tverven.github.io/tiai-seminar/ Organized together with Suraj Srinivas Tim van Erven

Should we trust LLM evaluations on publicly available benchmarks?🤔 Our latest work studies the overfitting of few-shot learning with GPT-4. with Harsha Nori Vanessa Rodrigues Besmira Nushi 💙💛 and Rich Caruana Paper: arxiv.org/abs/2404.06209 More details👇 [1/N]

📢 Excited to share our latest pre-print on evaluating & enhancing the safety of medical LLMs We introduce med-safety-benchmark to assess the #safety of #medical #LLMs and find that state-of-the-art models violate principles of medical safety and ethics. arxiv.org/pdf/2403.03744

[LG] Attention Mechanisms Don't Learn Additive Models: Rethinking Feature Importance for Transformers arxiv.org/abs/2405.13536 - Common transformer architectures like BERT and GPT-2 structurally cannot represent additive models like linear models or generalized additive

🧑‍🎨 Are Large Language Models Good Thieves? 🦹 Breakthroughs are not 0-to-1 processes. E.g. Picasso and the “invention” of cubism. Cubism was not a flash of thought⚡️ but developed gradually & used various 'inspiration' thefts (Juan Gris' work and the surrealist movement. To me,

🧵New paper: Machine Unlearning Fails to Remove Data Poisoning Attacks, ft Martin Pawelczyk (on Job Market), Jimmy Di, Ayush Sekhari (on Job Market), Seth Neel. Title says it all: current approaches for machine unlearning (MUL) are not effective at removing the effect of data poisoning attacks. 1/n

Can current unlearning methods remove poisoned training data from a trained model? Our new paper shows that unlearning methods are not quite ready, yet. Looking forward to a lot of interesting work ahead in this area.

Happy to share that our work on machine unlearning received a spotlight talk The GenLaw Center ICML workshop (1:30-2 pm) at Lehar 2. I will also chat about unlearning during the poster session (2-3 pm).

Wanna do a #PhD in #trustworthy #ML? Our group (sprintml.com) has 2 full-time positions open. If you are interested, please reach out. Please also share with people who might be interested.

AI subfields are evolving quickly over time. Some accelerating, while some are slowing down. #ResearchTrendAI More at researchtrend.ai

I will be hiring PhD students via ELLIS on multi-objective ML, causal generative models, and fair and interpretable ML. If interested, apply to our probabilistic ML: machinelearning.uni-saarland.de

📢I’ll be admitting multiple PhD students this winter to Columbia University 🏙️ in the most exciting city in the world! If you are interested in dissecting modern deep learning systems to probe how they work, advancing AI safety, or automating data science, apply to my group.

On this beautiful occasion of Diwali 🪔, I am happy to announce that our lab will recruit 1-2 PhD students in UVA School of Data Science for Fall 2025! Our group works on developing Scalable XAI and Trustworthy Algorithms for AI Alignment and Safety. Visit chirag-agarwall.github.io for

🚨📢 Excited to announce the ICLR 2025 Workshop on Building Trust in LLMs and LLM Applications! 📢🚨 Submit all your papers, and we’ll see you in Singapore! There will be paper awards, and we have a stacked lineup of speakers and panelists.

📢📢 Happy to share that our paper on unlearning evaluations has been accepted to ICLR 2025 🇸🇬 📜: arxiv.org/abs/2406.17216 Thanks to my great co-authors Seth Neel Gautam Kamath Jimmy Di Ayush Sekhari @yiweilu

Exciting opportunity at the intersection of climate science and XAI to work on groundbreaking research in attributing extreme precipitation events with multimodal models. Check out the details and help spread the word! #ClimateAI #Postdoc #UVA #Hiring Job description:

🚨 One Day Left! 🚨 Submit your work to the ICLR 2025 Workshop on Building Trust in LLMs and LLM Applications. We accept both Full Papers and Tiny Papers! building-trust-in-llms.github.io/iclr-workshop/…

📢 Last Chance! 📢 Submit your work to the ICLR 2025 workshop on Building Trust in LLMs & LLM Applications. ⏰ Deadline: 10 February AOE 📜 Call for papers: building-trust-in-llms.github.io/iclr-workshop/…