🚨𝐶𝑎𝑛 𝑎 𝑉𝐿𝑀 𝑤𝑎𝑟𝑛 𝑦𝑜𝑢 𝑎𝑏𝑜𝑢𝑡 𝑎 𝑤𝑒𝑡 𝑓𝑙𝑜𝑜𝑟 𝑤𝑖𝑡ℎ𝑜𝑢𝑡 𝑎 𝑠𝑖𝑔𝑛? or 𝑖𝑛𝑓𝑒𝑟 𝑡ℎ𝑎𝑡 𝑎 𝑏𝑎𝑏𝑦 𝑖𝑠 𝑐𝑟𝑦𝑖𝑛𝑔 𝑑𝑢𝑒 𝑡𝑜 𝑎 𝑚𝑖𝑠𝑠𝑖𝑛𝑔 𝑝𝑎𝑐𝑖𝑓𝑖𝑒𝑟? Introducing NL-Eye, a benchmark for assessing visual abductive reasoning in VLMs. >>

🚀 Excited to share our research: "Are LLMs Better than Reported? Detecting Label Errors and Mitigating Their Effect on Model Performance" 👇 arxiv.org/abs/2410.18889 #NLP #LLMs #MachineLearning

Breaking ReAct Agents: Foot-in-the-Door Attack 🚪🧑‍💻⚠️ 🚨 New preprint! 🚨 FITD attack subtly misleads LLM agents 🧠. Our reflection-based defenses counter these vulnerabilities! paper website: ibm.github.io/fitd-website/ with George Kour , Guy Uziel and Ateret Anaby-Tavor ⬇️ 1/11

שעות הלילה המאוחרות זה אחרי חצות, נכון? זה קריטי להימור ״באיזה חודש הפועל תל אביב תתחיל להתפרק״

🎉 Our paper "𝐍𝐚𝐯𝐢𝐠𝐚𝐭𝐢𝐧𝐠 𝐂𝐮𝐥𝐭𝐮𝐫𝐚𝐥 𝐂𝐡𝐚𝐬𝐦𝐬: 𝐄𝐱𝐩𝐥𝐨𝐫𝐢𝐧𝐠 𝐚𝐧𝐝 𝐔𝐧𝐥𝐨𝐜𝐤𝐢𝐧𝐠 𝐭𝐡𝐞 𝐂𝐮𝐥𝐭𝐮𝐫𝐚𝐥 𝐏𝐎𝐕 𝐨𝐟 𝐓𝐞𝐱𝐭-𝐭𝐨-𝐈𝐦𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬" is accepted to 𝐓𝐀𝐂𝐋! Eyal Ben-David Roi Reichart Anna Korhonen arxiv.org/abs/2310.01929 1/11

Our paper "On the Convergence of No-Regret Dynamics in Information Retrieval Games with Proportional Ranking Functions" has been accepted to #ICLR25! Joint work with the amazing team of Idan Pipano, Itamar Reinman and Moshe Tennenholtz. 🧵

Our paper: "On Behalf of the Stakeholders: Trends in NLP Model Interpretability in the Era of LLMs" has been accepted to NAACL 2025 main 🎉 We’ve updated the final version: 🔗arxiv.org/abs/2407.19200 If you are an NLP interpretability researcher, we have key takeaways for you!👇

1/🧵 New preprint: AdaptiVocab 📢 A lightweight method to make LLMs ~25% faster in domain-specific settings—without compromising quality. We adapt the vocabulary on top of your tokenizer to fit the domain. No architecture changes. No retraining from scratch.

Before I tell you about our new paper (now on arXiv), a quick question: 🧠 If I gave you a short text, could you guess whether the speaker has dementia? And how do you think GPT would do?

In a world dominated by GenAI, we forgot about the foundations of Machine Learning. It might sound crazy, but it's 2025 and we still use decision trees for the most fundamental tasks. We present TabSTAR, a Foundation Tabular Model🌟Eilam Shapira Roi Reichart arxiv.org/abs/2505.18125

אהלן! אני לקראת סיום התואר השני שלי במדעי הנתונים בטכניון, ומחפש עבודה כ-Data Scientist. יש לי רקע מחקרי ב-NLP ומדעי הקוגניציה, ניסיון בלמידה עמוקה ו-LLMים (כולל Fine-tuning ו-Prompt Engineering), ועבודה נרחבת עם חבילות Python רלוונטיות. אשמח לחיבורים וריטוויטים! הדיאם פתוח 🙏