Researcher @MetaAI AI - Neuroscience https://xitter.cf/t.co/VCaLTUt9MH
Paris, France
Joined March 2010
- Tweets1.2K
- Following531
- Followers8.4K
- Likes2.7K
Pinned Tweet
We’re happy to announce 2 releases today:
- 🧠Brain2qwerty v1 is published at @NatureNeuro
- 🚀 Brain2Qwerty v2 is now publicly released
Explore how we decode sentences from non-invasive brain recordings: facebookresearch.github.io/b…
Links:
📄v1 Nature Neuro: nature.com/articles/s41593-0…
📑v2 Meta preprint: facebookresearch.github.io/b…
💻Code: github.com/facebookresearch/…
📊Data: huggingface.co/datasets/bcbl…
📝Blog: ai.meta.com/blog/brain2qwert…
🧵Thread: xitter.cf/JarodLevy/status/20715…
Jean-Rémi King retweeted
A multilingual LLM encodes a sentence and its translation almost identically - the “platonic representation hypothesis.” Observed, never explained. The facts pile up: similarity peaks in the middle layers (not the ends), coexists with language-specific structure, and grows with language proximity, model quality, and data. We explain all of them -and predict something new: remove from each layer the part linearly predictable from the previous one, and cross-lingual similarity rises. Confirmed in real LLMs.
Jean-Rémi King retweeted
Excited to share that our paper “Polar probe linearly decodes semantic structures from LLMs” was accepted at COLM 2026! 🎉
Joint work with the amazing: @PierreOrhan, Emmanuel Chemla, @lakretz and @JeanRemiKing
We find a simple geometry for semantic structures in LLMs
👇🧵1/5
🚨We have a new position for a 3 year PhD at FAIR (Meta) Paris starting this fall/winter:
metacareers.com/profile/job_…
Looking for a candidate with strong ML/CS/AI background and a deep interest in computational neuroscience.
(Please RT)
Two major announcements for the #DigitalBrainProject:
🥇9 teams were selected in the first round
🚀we’re scaling the funding to $15M
More details below 👇:
digitalbrainproject.org
More details about the program and how to participate will be announced soon.
Press release: fo-rothschild.fr/sites/defau…
Project website:digitalbrainproject.org
🙏Thanks to the Rothschild Foundation Hospital (project management), the University of Montreal (project advising), @AIatMeta (funding) our scientific committee (Lune Bellec , Anne-Marie Kermarrec, Arthur Mensch, Russell Poldrack, Lucia Melloni, Jose-Alain Sahel), and our executive committee (Pierre Bourdillon, Elisabetta Simonetta, Julie Boyle, Elisa Escardi and myself), as well as to @AIatMeta for the trust and support !
Excellent investigation of the mechanics of language models.
🫢A notable blindspot, however: several teams have actually been **directly** comparing the working of LLMs to those of the human brain 🧠 for a while.
Here is a thread to highlight some of our (and our competitors'!) findings 👇
8. This is far from an exhaustive list. Many more work exist on the interpretability of LLM , with and without brain-comparisons e.g. @tallinzen @lakretz @roger_p_levy @ev_fedorenko @chrplr @alex_ander @martin_schrimpf @akgokce0 @mtoneva1 @tommmitchell
Does this mean that this work from @anthropic is incorrect? Absolutely not.
Does the global workspace theory applies to language ? Probably. But this theory does **not** mention language, and primarily focuses on perception.
It's seems to me all the more important to thus credit the work of those who have actually systematically compared brains and language models.