📢 We are delighted to announce that the 14th Lisbon Machine Learning Summer School will take place from the 11th to 17th of July 2024 Técnico Lisboa

🔗 for more information: lxmls.it.pt/2024/

🔗 to apply: tinyurl.com/288raakv

Deadline to apply: 26th April

LxMLS 2024 - 14th Lisbon Machine Learning School will be held this summer from July 11th to 17th, 2024. The call for participation is open until April 26, 2024.

🚩 Don't miss it, add it to your agenda! - lxmls.it.pt/2024/

#LxMLS2024
#MachineLearning
#NLP

Super excited about this new work!

Empirical: although LLMs have good abilities to generate questions, they don’t inherently know what’s important. We try to solve this!

Linguistic: is reader expectation predictable and if so, how well does that align with what’s in the text?

Excited to share a preview of Llama3, including the release of an 8B and 70B (82 MMLU, should be the best open weights model!), and preliminary results for a 405B model (still training, but already competitive with GPT4). Lots more still to come... ai.meta.com/blog/meta-llam…

How to enjoy the best of both worlds of efficient training (less communication and computation) and inference (constant KV-cache)?

We introduce a new efficient architecture for long-context modeling – Megalodon that supports unlimited context length. In a controlled head-to-head

Looking for a unique low-resource setting? Tired of evaluating on the same old benchmarks? Look no further! Thanks to Heather Lent & a team of collaborators, we are proud to release
✨CreoleVal ✨- a collection of benchmarks for 8 Creole languages #NLProc arxiv.org/abs/2310.19567

4th Advanced Language Processing Winter School #ALPS24 starts Mon in breathtaking Vanoise (FR alps)😏! Fab speakers Sara Hooker Barbara Plank Thomas Wolf Andre Martins Claire Gardent Maxime Peyrard & jolly organisers LIG NAVER LABS Europe cohere Info: lig-alps.imag.fr

Help us break LLMs! The test suite sub-task will be included for the sixth time in the General MT Shared Task of the Conference on Machine Translation (WMT24). This year's theme is to reveal weaknesses of LLMs when translating #LLMs #EMNLP #NLProc | www2.statmt.org/wmt24/testsuit…

Are you attending the eaclmeeting? Don't miss Chryssa Zerva's keynote talk on 'Uncertainty in NLP: Quantification, interpretation and evaluation' at the UncertaiNLP workshop on the 22nd March.

🔗 more info uncertainlp.github.io/program

The Chat Shared Task (WMT2024) is live! 💥💥

Happy to announce this year’s Chat Shared Task which aims to translate a corpus composed of genuine bilingual conversations from the customer support domain!

Glad to share that our AfriCOMET paper has been accepted at #NAACL2024 . See you in Mexico.

Try out our model on Hugging Face

huggingface.co/models?sort=tr…

with Jiayi Wang, Sweta Agrawal Ricardo Rei Eleftheria Briakou Marine Carpuat
Xuanli He
Masakhane

Today we release the Tower paper! 🗼
Tower is an open-weight suite of multilingual models — built on top of LLaMA-2 — for translation-related tasks. It supports 10 different languages.

Paper: arxiv.org/pdf/2402.17733…
Models and data: huggingface.co/collections/Un…
🧵Thread below.

Check out work from Sebastian Joseph and Lily Chen, collaborating with medical experts, on the intricate task of factuality eval of LLM summarization and simplification for evidence-based medicine! FactPICO opens up new avenue for expert-level explainable factuality benchmarking

🎉 Our great team has just released a much improved Tower! We reach super high performance with TowerInstruct-13B, particularly for MT, outperforming much bigger models and dedicated translation models.

Next step: beating GPT-4? 👀

Bonus news: the paper is coming soon! 👨🏻‍🍳

CroissantLLM 🥐

A Truly Bilingual French-English Language Model

paper page: huggingface.co/papers/2402.00…

introduce CroissantLLM, a 1.3B language model pretrained on a set of 3T English and French tokens, to bring to the research and industrial community a high-performance, fully

Very excited about this work! Conformal NLG that also accounts for the non-exchangeability of text generation!

#EACL2024 Findings

I'm excited to share that our paper 'Non-Exchangeable Conformal Risk Control' (with Chryssa Zerva Dennis Ulmer (is on the job market 👨🏻‍💻) Andre Martins) has been accepted at #ICLR2024 . Check out the updated version of the paper: arxiv.org/abs/2310.01262. See you in Vienna!

Introducing Tower our cutting-edge multilingual #LLM for translation-related tasks! 🚀
With 7B parameters and support for 10 languages, Tower dominates in pre-translation tasks and machine translation. 🌎
Explore the future of #NLP now 👉 hubs.li/Q02g7_9B0

What if I told you that you can simultaneously enhance an LLM's task performance and reduce its size with no additional training?

We find selective low-rank reduction of matrices in a transformer can improve its performance on language understanding tasks, at times by 30% pts!🧵