Sahar Abdelnabi 🕊 (on 🦋)
@sahar_abdelnabi
Researcher @ Microsoft | ex. PhD @ CISPA | Neurodivergent 🧠🦋 | AI safety & security | life and peace for all ☮️, permanent ceasefire 🍉 Opinions my own.
Hawthorne effect describes how study participants modify their behavior if they know they are being observed In our paper 📢, we study if LLMs exhibit analogous patterns🧠 Spoiler: they do⚠️ 🧵1/n
Deadline extended to July 3rd!
🚀 Exciting Announcement! 🚀 Get ready for the 18th ACM Workshop on Artificial Intelligence and Security (AISec 2025)! 📍Co-located: @acm_ccs 🗓️ Deadline: June 20th, 2025 🌐 Website: aisec.cc w/ @ruoxijia and Matthew Jagielski
NeurIPS is pleased to officially endorse EurIPS, an independently-organized meeting taking place in Copenhagen this year, which will offer researchers an opportunity to additionally present their accepted NeurIPS work in Europe, concurrently with NeurIPS. Read more in our blog…
There are many great researchers out there. But the ones that really stand out to me are the ones who are also kind, even when they don't need to be.
Join our mission to strengthen AI research in Europe 🇪🇺 We are looking for several ML Research Engineers and Scientists to work on OpenEuroLLM at the ELLIS Institute Tübingen. If you're passionate about large-scale model training, multilingual evaluation and want to contribute to…
We are very happy that our paper was selected as *Oral* and for a spot in the Panel Discussions (among only 25 papers) at #ACL2025NLP!🏆🥳 Our work sheds light on how LLMs sample their responses from vast possibilities. Have a read below if you are interested to know more.
📢📢Our paper, "A Theory of Response Sampling in LLMs: Part Descriptive and Part Prescriptive" has been accepted at ACL 2025 (Main Conference)!! 🥳 Details below 👇🧵1/n
You cannot oppose the killing of civilians in one place and justify it in another. If your stance varies based on who the victims are, it is not a moral position—it’s a selective, morally bankrupt double standard.
Hopefully my only tweet on this. We see wars as locations and numbers since we don’t get to know the people — read their story and picture their persona. This takes 6 minutes from you and depicts a real view of the ones that are under fire these days. medium.com/@soroushzargar…
In war, both sides lose. That we don’t learn this is the greatest tragedy.
Honored to be invited to the Graz security week summer school! Looking forward to the event ☺️
Join the Graz Security Week from Sep 1 to 5! with @sahar_abdelnabi, @jovanbulck, Maria Eichlseder, Georg Fuchsbauer, @sublevado, @fbpierazzi, @kavehrazavi, @chrossow, @realyangzhang on topics system security, side channels, AI Security, and Cryptography: securityweek.at
This looks like a really interesting approach to how models respond to the knowledge they are being evaluated.
Hawthorne effect describes how study participants modify their behavior if they know they are being observed In our paper 📢, we study if LLMs exhibit analogous patterns🧠 Spoiler: they do⚠️ 🧵1/n
Contributed with hundreds of attacks in this dataset, it was a great challenge. Maybe i will make a second part of m19o.github.io/posts/Phishing…
📢 We are releasing the dataset and write-up for the LLMail-Inject challenge! We ran a public challenge simulating real-world prompt injection attacks against LLM-based email assistants where the attackers could adapt their attacks to each defense 1/n 🧵 arxiv.org/abs/2506.09956