כמו כל מודל חדש בעברית Shaltiel ואני הרצנו כמה בדיקות על דאטה סטים קיימים בעברית. והתוצאות... קצת מוזרות. הבדיקות הן: וינוגרד - דאטה סט של common sense HeQ - דאטה סט של שאלות ותשובות סנטימנט אנחנו רואים ירידה משמעותית בתוצאות ביחס למודל המקורי של גוגל. המודל עובד לכם?

Our latest app update will bring sentence segmentation for better translation reliability with multiple sentences. It's a trade-off, improving reliability at the expense of context-awareness. Your feedback is crucial as we evolve. #SignLanguage

Just so y'all know: Every time you push code, you appear on this real-time wall at the GitHub HQ.

I wish everyone were the tiny corp

🐣Too Long; Didn't Read: Most LLMs cite full documents for support, but they're too long so people don't actually read them. In our new work "Attribute First, then Generate”, we introduce fine-grained attributions, where LLMs are required to highlight only relevant information

🚀 Dicta-LM 2.0 - מודל שפה גנרטיבי גדול בעברית, פתוח, וחזק! 🎉 דמו של גרסת הצ'אט: shorturl.at/BPY09 אתר החדש של דיקטה עבור תרגום מכונה אנגלית<->עברית שמבוסס על המודל: translate.dicta.org.il קישור לבלוג המלא: dicta.org.il/dicta-lm 🧵>>

🚀 Dicta-LM 2.0 - a large, open, and powerful generative language model in Hebrew! 🎉 A demo of the chat version: shorturl.at/BPY09 English<->Hebrew machine translation based on the model: translate.dicta.org.il Link to the full blog: dicta.org.il/dicta-lm

Thrilled to launch the Open LLM Leaderboard for Hebrew, as a continuation of the Dicta-LM 2.0 release! This creates a new set of benchmarks for evaluating Hebrew language models. Blog: huggingface.co/blog/leaderboa… Leaderboard: huggingface.co/spaces/hebrew-…

🎉 This weekend, we tackled a persistent bug in @google MediaPipe Holistic pose estimation. 🛠️ Enhanced the hand ROI predictions crucial for #SignLanguage processing—now with better accuracy! 🔍 Dive into our approach and help deliver it! arxiv.org/abs/2405.03545

Wanna fool yourself that GPT4o can understand sign language? Name your file with the translation, and your username as a part of the answer, and see how it works like magic!

🚨 New Paper 🚨 Are current LLMs up to the task of solving *complex* instructions based on content-rich text? Our new dataset, TACT, sheds some light on this challenge. How does it work? arxiv.org/abs/2406.03618 Work by Google AI & Google DeepMind 👇🧵

sign language 🤦‍♂️

#ACL2024 A belated highlights list: - Presenting our work and engaging in discussions - Bangkok with my wife and 1-year-old daughter - Amazing food - A much needed break from challenging times at home Grateful for the well-organized conference and the opportunity to connect!

Ansub sign.mt by Amit Moryossef

🐳Introducing Knowledge Navigator🐳: A new way to explore scientific literature! Our paper shows how LLMs can transform information overload into structured, navigable knowledge. paper: arxiv.org/pdf/2408.15836 page (with demo): knowledge-navigators.github.io 🧵/9

Announcing the **Demo Paper Awards** for #EMNLP2024 Best Demo Paper and Outstanding Demo Paper:

Thrilled to receive the outstanding demo award at #EMNLP2024 by EMNLP 2025 for our sign language translation demo 🥳

congratulations to Amit Moryossef for winning the outstanding demo paper award at #EMNLP2024!

I'm literally speechless.... 🤯 Gajus 🙏

Honoured to receive two (!!) Senior Area Chair awards at #ACL2025 😁 (Conveniently placed on the same slide!) With the amazing Philip Whittington, Gregor Bachmann and Ethan Gotlieb Wilcox, Cui Ding, Giovanni Acampa, Alex Warstadt, Tamar Regev