Wanna check how well a model can share knowledge between languages? Of course you do! 🤩 But can you do it without access to the model’s weights? Now you can with ECLeKTic 🤯

Producing reasoning texts boosts the capabilities of AI models, but do we humans correctly understand these texts? Our latest research suggests that we do not. This highlights a new angle on the "Are they transparent?" debate: they might be, but we misinterpret them. 🧵

@rtsarfaty It was fun :D

First contributed talks are under way for Phonology, Morphology, and Syntax! Hall M1 Level 1. Schedule here: https://t.co/c3BDfhP605 #ACL2025NLP #CoNLL2025

🚨 New paper alert! 🚨 We propose an IQ Test for LLMs — a new way to evaluate models that goes beyond benchmarks and uncovers their core skills. Think: 🧠🤖 psychometrics for LLMs. 👇 (1/6)

Are you still around Vienna? Come hear about a new morphological task at CoNLL at ~11:20 (hall M.1) @rtsarfaty

whlie you drink you coffee at 10am, stop by the google booth to talk about cross-lingual transfer 🤓 https://t.co/KQsRC86ey1

Will be especially happy to talk to UK-based people as I'm moving in a couple of months to @CambridgeLTL

On my way to #ACL2025 ! 🤗 Find me talking about crosslingual transfer at the Google booth, morphology - at @conll_conf , and tokenization - just at the coffee breaks

🚨 RAG is a popular approach but what happens when the retrieved sources provide conflicting information?🤔 We're excited to introduce our paper: “DRAGged into CONFLICTS: Detecting and Addressing Conflicting Sources in Search-Augmented LLMs”🚀 A thread 🧵👇

🤯 MIND-BLOWN! A new paper just SHATTERED everything we thought we knew about AI reasoning! This is paradigm-shifting. A MUST-READ. Full breakdown below 👇 🧵 1/23

I really wanted to see the review details. It's clearly above the acceptance threshold of findings for me. When you fall into the cycle of rejection from ARR, it's hard to come out.

https://t.co/nuXZFrQbf7

RefVNLI Towards Scalable Evaluation of Subject-driven Text-to-image Generation

English is full of “lexical gaps,” words that are implied to exist but don’t, because we borrowed a bunch of words from Latin but not other, related words. Somebody made a chart to show it↓

🚨 NEW WORKSHOP ALERT 🚨 We're thrilled to announce the first-ever Tokenization Workshop (TokShop) at #ICML2025 @icmlconf! 🎉 Submissions are open for work on tokenization across all areas of machine learning. 📅 Submission deadline: May 30, 2025 🔗

here is it in a thread form:

our research featured in a Google Research blog post! @Uri_Shaham @mataneyal1

I really enjoyed working on this paper! Towards more challenging multilingual research questions!

Evaluate & improve your favorite model with ECLeKTic! 📜paper: https://t.co/uzbDbFzCNj 📊data: https://t.co/pyTz8SkrYA @Uri_Shaham Dan Malkin @mataneyal1 Sivan Eiger Avinatan Hassidim @ymatias @maynez_joshua @AdiMayrav @jasonriesa @shrutirij @laurarimell Idan Szpektor @rtsarfaty

In line with previous works, we also found some evidence that a shared script eases the transfer of knowledge in ECLeKTic. Note that Indonesian is part of the “Latin script” block although it’s not genealogically related