Antoine Bosselut
@ABosselut
Followers
4K
Following
2K
Media
48
Statuses
1K
Helping machines make sense of the world. Asst Prof @ICepfl; Before: @stanfordnlp @allen_ai @uwnlp @MSFTResearch #NLProc #AI
Joined March 2013
The next generation of open LLMs should be inclusive, compliant, and multilingual by design. That’s why we (@EPFL @ETH_en) built Apertus.
@EPFL , @ETH_en and #CSCS today released Apertus, Switzerland's first large-scale, multilingual language model (LLM). As a fully open LLM, it serves as a building block for developers and organizations to create their own applications: https://t.co/7bJlINiIdn
#Apertus #AI
2
21
106
Looking forward to speaking at cAIroNLP next Thursday (October 23) about our recent work on building modular language models with brain-like specialization. Hope to see you there! Event Link: https://t.co/WentsJRxdA
0
3
11
Congratulations to the EPFL team selected for the AI for Math Fund by @RenPhil21 with support from @xtxmarkets! 🎉 Their project, Document-Level Autoformalization, uses AI to bridge human and machine understanding of mathematics. 🔗 Learn more: https://t.co/MbtAJmmgwR
@ICepfl
0
4
5
If you're interested in doing a postdoc at @ICepfl there's still time to apply for the @EPFL_AI_Center postdoctoral fellowships. Apart from this, I'm also recruiting postdocs in developing novel training algorithms for reasoning models and agentic AI.
1
6
25
Join us again at #MELT workshop at #COLM2025 to hear from @ImanolSchlag about #Apertus, the largest multilingual LLM trained on over 1000 languages.
0
0
7
Kicking off #MELT workshop at #COLM2025 with @monojitchou talking about "Meta-Cultural Competence: What LLMs Should Know About Culture to Serve the Next Billion Users" !
0
2
19
Come join us in 520D (all the way down the hall and around the corner) at #COLM2025 for the first workshop on multilingual and equitable language technologies!
0
2
11
Introducing ModernVBERT: a vision-language encoder that matches the performance of models 10× its size on visual document retrieval tasks! 👁️ Read more in the thread👇 (1/N)
7
33
209
In case you haven't seen, the updated report is finally on arXiv: https://t.co/5v38cljnW0 And if you believe that such an initiative is cool: we are hiring research engineers! Come and work with the biggest public AI supercomputer --- 10k GPUs :) Link: https://t.co/yIpvTPOzEX
Long in the making, finally released: Apertus-8B and Apertus-70B, trained on 15T tokens of open data from over 1800 languages. Unique opportunity in academia to work on and train LLMs across the full-stack. We managed to pull off a pretraining run with some fun innovations, ...
2
8
47
I’ve recently joined @Alibaba_Qwen! We’re building the next generation of frontier models through careful science and world-class engineering, and we are making rapid progress. Excited for what’s ahead 💜
62
23
913
We’re kicking off the day at the Swiss {ai} Hackathon in Lausanne, co-organized by @LauzHack and @epflaiteam. The morning starts with a short talk by @YohannPERRON from @Innovaud, followed by a presentation of @apertus_ai by Martin Jaggi. #AI #Hackathon
1
2
6
Check out Deniz's work on using crosscoders to uncover emergence and consolidation of linguistic features in LLMs across pretraining checkpoints.
1/🚨 New preprint How do #LLMs’ inner features change as they train? Using #crosscoders + a new causal metric, we map when features appear, strengthen, or fade across checkpoints—opening a new lens on training dynamics beyond loss curves & benchmarks. #interpretability
0
1
13
💡Can we optimize LLMs to be more creative? Introducing Creative Preference Optimization (CrPO) and MuCE (Multi-task Creativity Evaluation Dataset). Result: More novel, diverse, surprising text—without losing quality! 📝 Appearing at #EMNLP2025
1
6
13
For anyone at @EPFL: I’m co-organizing this mini-symposium on Evaluations, Impact, and the Science of AI Safety on Sep 29 w/ @ABosselut & @EPFL_AI_Center. Will be a great way to meet & hear from researchers from @AISecurityInst. See you in Lausanne! 🇨🇭🇬🇧 https://t.co/IVUqSxGtwN
4
2
13
Today we're announcing Gauss, our first autoformalization agent that just completed Terry Tao & Alex Kontorovich's Strong Prime Number Theorem project in 3 weeks—an effort that took human experts 18+ months of partial progress.
82
487
3K
XLM-R has been SOTA for 6 years for multilingual encoders. That's an eternity in AI 🤯 Time for an upgrade. Introducing mmBERT: 2-4x faster than previous models ⚡ while even beating o3 and Gemini 2.5 Pro 🔥 + open models & training data - try it now! How did we do it? 🧵
13
64
249
3T tokens, ~1800 languages, 2 models - we’re releasing mmBERT, a modern multilingual encoder model!
9
65
400
Proud to have been part of this endeavor! 🚀
🚀 Together with ETH Zürich and the CSCS, we have just released Apertus, 🇨🇭 Switzerland’s first large-scale, open, multilingual language model — a milestone in generative AI for transparency and diversity. Find out more:
1
3
29
A big step for Switzerland 🇨🇭 and a great achievement for our in-house alignment algorithm QRPO ( https://t.co/qfNVcjpm0X) which has shown remarkable stability and predictability at the 70B scale 🚀!
Long in the making, finally released: Apertus-8B and Apertus-70B, trained on 15T tokens of open data from over 1800 languages. Unique opportunity in academia to work on and train LLMs across the full-stack. We managed to pull off a pretraining run with some fun innovations, ...
1
5
20
(5) Transparency: We're fully open, pairing our weights with a full suite of reproduction artifacts. Check out our artifacts and technical report here:
huggingface.co
1
1
5