José G. C. de Souza

3.5K posts

José G. C. de Souza banner
José G. C. de Souza

José G. C. de Souza

@accezz

PhD, Computer Science, Multilingual NLP and LLMs, Generative AI, Machine Translation, Machine Learning | Previously @eBay, PhD @UniTrento / @FBK_research

Lisbon, Portugal Katılım Nisan 2008
1K Takip Edilen500 Takipçiler
José G. C. de Souza retweetledi
Pedro Martins
Pedro Martins@PedroHenMartins·
Today we release EuroLLM-9B: the best EU-made multilingual LLM of its size! Check the blog post for more info and results: huggingface.co/blog/eurollm-t…. Stay tuned for the technical report and bigger and more powerful models!
English
4
24
87
13.8K
José G. C. de Souza
José G. C. de Souza@accezz·
Super happy with this new product based on our Tower models we have been developing in the last year! Try it out and let us know what you think and what could be improved! wind.ai
Unbabel@Unbabel

💥 Today we’re excited to announce the launch of hubs.li/Q02Y2GpL0 - our new standalone AI solution built for businesses looking to scale quickly with cost-effective translations you can trust. 👇 Learn more about Widn and try it for free. hubs.li/Q02Y2G4q0

English
0
0
4
178
José G. C. de Souza retweetledi
Pedro Martins
Pedro Martins@PedroHenMartins·
Today we release the first EuroLLM paper and models: EuroLLM-1.7B and EuroLLM-1.7B-Instruct! The EuroLLM project will develop open-weight multilingual LLMs that understand and generate text in all official EU languages. Stay tuned for the bigger and stronger EuroLLMs (9B, 22B)!
English
3
18
77
13.4K
José G. C. de Souza retweetledi
Nuno M. Guerreiro
Nuno M. Guerreiro@nunonmg·
The good news continue 🥳 We are delighted that Tower has been selected as a oral spotlight presentation at @COLM_conf (top 2% of submissions)! Check out our paper here: openreview.net/pdf?id=EHPns3h…
José Maria Pombal@zmprcp

I’m super proud to announce that our Tower 🗼 paper was accepted at COLM 24! To celebrate, we are releasing a new version of TowerInstruct based on Mistral 7B: it achieves similar performance to our 13B model, while being almost half the size! COLM Paper: openreview.net/pdf?id=EHPns3h…

English
3
11
46
7.7K
José G. C. de Souza retweetledi
José Maria Pombal
José Maria Pombal@zmprcp·
I’m super proud to announce that our Tower 🗼 paper was accepted at COLM 24! To celebrate, we are releasing a new version of TowerInstruct based on Mistral 7B: it achieves similar performance to our 13B model, while being almost half the size! COLM Paper: openreview.net/pdf?id=EHPns3h…
José Maria Pombal tweet media
English
1
13
48
11.4K
José G. C. de Souza retweetledi
laurent besacier
laurent besacier@laurent_besacie·
We offer this 1y postdoc to work with us on the @UTTERProject EU project on LLM-based agents ! Come work with us on 1 or several of these topics: i] managing uncertainty and ambiguity ii] improving the use of conversational context iii] ensuring the safety and alignment of LLMs.
NAVER LABS Europe@naverlabseurope

📢 Open position! PostDoc position in #LLM powered conversational agents @naverlabseurope Grenoble, France. @UTTERProject ***Please share*** Start date: September Duration: 1yr More info & how to apply: europe.naverlabs.com/job/postdoc-ll…

English
0
8
13
1.6K
José G. C. de Souza retweetledi
Wafaa
Wafaa@Wafaa01997·
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!
English
1
9
13
1.8K
José G. C. de Souza retweetledi
Nuno M. Guerreiro
Nuno M. Guerreiro@nunonmg·
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.
Nuno M. Guerreiro tweet media
English
4
46
149
14.7K
José G. C. de Souza retweetledi
Nuno M. Guerreiro
Nuno M. Guerreiro@nunonmg·
🎉 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! 👨🏻‍🍳
Unbabel@Unbabel

🚀 Exciting news! Our new TowerInstruct-13B is the top open-weight model for translation tasks, outperforming competitors and even challenging closed models like GPT-3.5 and GPT-4. Explore Tower's enhanced capabilities now 👉 hubs.li/Q02kJGMz0

English
1
7
25
2.3K
José G. C. de Souza
José G. C. de Souza@accezz·
We are releasing: * TowerBase, a continued pre-trained LLaMA2 * TowerInstruct, a finetuned TowerBase on a curated instruction set * TowerBlocks, the instruction set used for TowerInstruct huggingface.co/collections/Un…
English
2
0
1
119
José G. C. de Souza
José G. C. de Souza@accezz·
Super happy to share something we have been working on lately: TowerLLM, a multilingual model geared towards cross-lingual and translation-related tasks. It has very good performance on translation benchmarks and supports 10 languages.
Unbabel@Unbabel

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

English
1
2
36
2.2K
José G. C. de Souza retweetledi
Nuno M. Guerreiro
Nuno M. Guerreiro@nunonmg·
I am thrilled to share xCOMET! 😇 This metric has been in the works for a while and we are super proud of it. xCOMET will provide, contrary to previous efforts, sentence-level scores alongside error spans and their severity. Check the paper: 📑 arxiv.org/abs/2310.10482
Nuno M. Guerreiro tweet media
English
3
15
60
9.7K
José G. C. de Souza retweetledi
Unbabel
Unbabel@Unbabel·
Excited to share that we've won the prestigious #QualityEstimation shared task at #WMT23 for a 2nd year in a row!🎉 Our winning build was a LLM QE system with 11B parameters -- the largest #QE system ever built. 💪 Check out the win here >> hubs.li/Q021rMgZ0 @machtranslate
English
0
1
6
1.2K
José G. C. de Souza retweetledi
Darcey Masters (née Riley)
Darcey Masters (née Riley)@DarceyNLP·
Old-ish news but this paper is *so good*. It does such a thorough and systematic comparison of, like, every conceivable setting you could use for n-best reranking or MBR. Extremely useful for figuring out what actually works.
Patrick Fernandes@psanfernandes

Tired of beam search and all the heuristics needed to make it work well in MT? In our work accepted at #NAACL2022 (co-lead @tozefarinhas) we explore an alternative decoding method that leverages neural metrics to produce better translations! arxiv.org/abs/2205.00978 1/14

English
0
4
18
5.6K
José G. C. de Souza retweetledi
Ricardo Rei
Ricardo Rei@RicardoRei7·
Improving Machine Translation Evaluation with COMET v2.0 - Say Goodbye to Outdated Metrics! 🌟 With all the hype around GPT-4 is more important than ever to have reliable evaluation metrics and methodologies. That why we are excited to release COMET v2.0: resources.unbabel.com/r-d-blog/intro…
English
4
15
39
3K
José G. C. de Souza retweetledi
Paolo Crosetto
Paolo Crosetto@PaoloCrosetto·
Does your (EU) country attract or lose top researchers? AT, CH attract a lot. FR, DE, BE, ES are open places: as many leave as arrive. IE is a closed system. Italy, there is something very wrong: *very* few come, many go. Plot made with @ERC_Research data.
Paolo Crosetto tweet media
English
66
1.2K
4.3K
0

Keşfet

@swetaagrawal20 @tozefarinhas @RicardoRei7 @andre_t_martins @COLM_conf @UTTERProject @duartealves @nunonmg