Sign in

Léna Gaubert

@kittog.bsky.social
73 followers 150 following 131 posts

nlp (among other things) 🎪 data scientist @ arlq.ai 🔗 kittog.github.io

PostsRepliesMedia
Reposted by Léna Gaubert
Shreya Chowdhary @shreyachowdhary.bsky.social · 08/10/2026
New #AIES paper w/ @benzevgreen.bsky.social & @tisjune.bsky.social - "Deconstructing AI Hype: A Playbook for Resistance"! Read here: shreyachowdhary.net/files/decons... We argue that AI hype is a problem not because it misdescribes AI, but because it facilitates AI entrenchment.
12715
Reposted by Léna Gaubert
Phillip Isola @phillipisola.bsky.social · 09/10/2026
In PRH, we argued that rep geometry is converging, but it's still an open question exactly in what sense. I want to share some new evidence that might clarify the picture. The evidence comes from our work here, led by @schnaus.bsky.social: dominik-schnaus.github.io/unpaired-ros... 1/n
13912
Reposted by Léna Gaubert
CrimethInc. Ex-Workers' Collective @crimethinc.com · 06/10/2026
People in Ithaca sent a message to those who would normalize sexual assault on campus, paying a visit to Cornell’s main administrative building, Day Hall, which houses the office of the university president. If graffiti didn't change anything, it would be legal. crimethinc.com/KillPower
Day Hall at Cornell University vandalized early Monday morning, October 5, 2026, with graffiti reading "Smash the patriarchy" and "At Cornell, rapists are protected, protestors are expelled."
2751244
Reposted by Léna Gaubert
Mayor Zohran Kwame Mamdani @mayor.nyc.gov · 02/10/2026
Black text on a white that says the following:

The allegations in the Cornell sexual assault case are deeply disturbing and gut-wrenching. Cornell University has failed this survivor.

I fully support Governor Hochul's decision to name Attorney General James as special prosecutor in this case and look forward to the complete and extensive investigation that is to come. 

Accountability and justice cannot simply be values, they must also be delivered.
9284561466
Reposted by Léna Gaubert
ELLIS @ellis.eu · 01/10/2026
🚨 Apply now for the #ELLISPhD Program! 🎓 Work with leading AI/ML researchers across Europe, with exchange visits & joint supervision. The application deadline is 31 Oct. Get all info & apply now at ellis.eu/news/ellis-p...
077
Reposted by Léna Gaubert
Kate Manne @katemanne.bsky.social · 30/09/2026
I wrote for @thecut.com about the Cornell case and what it says that, at my institution and many others like it, protesters are often punished more harshly than sexual assailants www.thecut.com/article/what...
thecut.com
What It Feels Like at Cornell Right Now
As a professor at the Ivy League school, Kate Manne has known many, many victims. There’s a palpable feeling on campus, she writes, that this is big.
521873618
Reposted by Léna Gaubert
Anne Helen Petersen @annehelen.bsky.social · 30/09/2026
This is *the* piece. Kate Manne, professor of philosophy at Cornell, speaks truth to the totality of failures at the university — and why students protesting the war in Gaza have faced far harsher consequences than any of the Cornell 7: www.thecut.com/article/what...
thecut.com
What It Feels Like at Cornell Right Now
As a professor at the Ivy League school, Kate Manne has known many, many victims. There’s a palpable feeling on campus, she writes, that this is big.
5030971089
Reposted by Léna Gaubert
ali alkhatib @ali-alkhatib.com · 30/09/2026
shutting down frat houses won't end rape culture, but it will disorganize a lot of rapists, and i think that's probably worth doing if only to destabilize what is apparently an ecology of organizations that functionally exist to rape women per the whole "purpose of a system is what it does" insight.
84001761
Reposted by Léna Gaubert
ELLIS @ellis.eu · 25/09/2026
Hey NLP enthusiasists! Check out this event supported by the ELLIS Unit Denmark: a free one-day #NLProc symposium featuring international speakers, poster presentations, and networking activities. 📍 Copenhagen 🇩🇰 📅 22 October 🔗 cphnlp.github.io
0164
Léna Gaubert @kittog.bsky.social · 25/09/2026
I'm pretty I really only needed one notebook for the two months to come, not sure what happened here
000
Reposted by Léna Gaubert
Simon Fisher @profsimonfisher.bsky.social · 11/09/2026
If you're having a stressful day, why not sit back, relax & immerse yourself in the reconstructed soundscape of a Jurassic forest. Based on ingenious analyses of the fossilized wings of extinct insects, from a paper in latest issue of @pnas.org. www.pnas.org/doi/10.1073/... (Sound up, obviously.)🦗🔊🧪
26225
Reposted by Léna Gaubert
Laurent Besacier @lbesacier.bsky.social · 15/09/2026
We are running #ALPS again in 2027! 1 week in a stunning alpine location, awesome lectures about modern NLP, constant interactions with fellow students & researchers Applications opened today alps.imag.fr
032
Léna Gaubert @kittog.bsky.social · 30/08/2026
pretty please!! noaifridays.com
noaifridays.com
No AI Fridays
A weekly ritual for software teams to unplug from AI coding assistants, prevent skill atrophy, and rediscover the joy of craftsmanship.
020
Léna Gaubert @kittog.bsky.social · 18/08/2026
Great work from @geometric-intel.bsky.social in collaboration with Arlequin AI! Published at TAG-DS this week. 🎪 :)
022
Léna Gaubert @kittog.bsky.social · 10/08/2026
probably the coolest website that we could ever have for latent spaces
021
Léna Gaubert @kittog.bsky.social · 09/08/2026
Had forgotten some in another roll of expired Kodak Gold 200!
Cliff walk, Howth.Two highland cows, resting on grass under the sun.Cat sitting on gravel, looking at the camera. Some orange flowers grew through the gravel. Cat's fur is black, ginger and white.Howth's beach. At the back, cat from the previous picture, waiting.
000
Léna Gaubert @kittog.bsky.social · 05/08/2026
food for thoughts: - you can't fully judge something until you've used it (im critical of the practice itself and its impact in research & learning) - CLAUDE(.)me files exist to list guidelines and maintain control - you don't have to use it when you don't want to
000
Léna Gaubert @kittog.bsky.social · 05/08/2026
cons: - AI fatigue/context switching - potentially losing independence over generated code - Claude can be a (very) frustrating model - i don't want it in vscode (too distracting, maybe the CLI would work better?) - maybe learning less (?)
100
Léna Gaubert @kittog.bsky.social · 05/08/2026
pros: - help developing interface for experiences (front end) - faster dev for my CLI (i tend to prefer CLI over notebooks and what not) & help maintaining my repos up to date - reduce some frustrations encountered daily when coding (?) - assist in brainstorming
110
Léna Gaubert @kittog.bsky.social · 05/08/2026
slowly considering the idea of potentially giving in and using claude code after months of fighting the idea of having that thing in my ide. my mind has been spiraling over this for weeks now and it's so silly but i'm still atrociously conflicted.
110
Reposted by Léna Gaubert
dell cameron @dell.bsky.social · 05/08/2026
NEW: Meta reviewed, approved, and ran more than 50 ads containing AI-generated child sexual abuse material across Facebook, Instagram, Messenger and Threads. They ran for nine months. Some were live this week. After WIRED asked about them, researchers found ~30 more
wired.com
Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery
More than 50 offending image and video ads were published across Facebook, Instagram, Messenger, or Threads, according to Meta’s ad library data. Some ran as recently as this week.
20738445
Reposted by Léna Gaubert
Eryk Salvaggio @eryk.bsky.social · 05/08/2026
When you prompt a model that is designed to produce text in the shape of a plan, it should not surprise you when a plan emerges
0164
Léna Gaubert @kittog.bsky.social · 04/08/2026
had a great time presenting our work at #CogSci2026 during the human-AI interaction panel. this was a first for me, so I was definitely over-stressed until the time of the talk, but it was a great experience overall :) 🇧🇷
Me presenting! Slide on experiment design projected on big screen.View from the beach of the Sheraton Hotel.
000
Reposted by Léna Gaubert
Geometric Intelligence Lab @geometric-intel.bsky.social · 30/07/2026
🎉Congratulations to Guillermo Bernárdez Gil! Thrilled to share that our postdoc, Guille, has been awarded the 2026 Extraordinary Doctoral Award by the Universitat Politècnica de Catalunya (UPC). Congrats, Guille, so proud to have you on our team👏 🔗https://tinyurl.com/43hbr9t3
275
Reposted by Léna Gaubert
Jalal Al-Tamimi @beppefusilli.bsky.social · 24/07/2026
Dear all,, one of my M1 students is recruiting participants for two online listening experiments. We are looking for native speakers of English, aged 18–50, with no reported history of speech, hearing, or neurological disorders. Participation is entirely voluntary and unpaid.
122
Léna Gaubert @kittog.bsky.social · 24/07/2026
In this project, you'll be working with me alongside amazing researchers from INRIA (Serena Villata, Federica Granese, Romain Badouard). There will also be additional joint work with @norijacoby.bsky.social (CoCoCo Lab, Cornell). For additional info, feel free to reach out: lena@arlq.ai
000
Léna Gaubert @kittog.bsky.social · 24/07/2026
This project picks up our preliminary experiments on opinion diffusion in AI-human networks: here, we'd like to understand how fake news & misinformation propagate in these new hybrid collectives, via large-scale online social experiments. :)
100
Léna Gaubert @kittog.bsky.social · 24/07/2026
📢 Our wonderful R&D team here at Arlequin AI is looking for a Postdoc in Computational Social Science to join a fully-funded project in collaboration with MARIANNE & SIRA teams at INRIA. 📆 18 months 📍Paris, France
110
Léna Gaubert @kittog.bsky.social · 22/07/2026
Super excited to finally be in Rio for #CogSci2026!! 🇧🇷 :) I'll be presenting our paper at the Human-AI interaction panel this Friday morning. Looking forward to great discussions all week long!
052
Reposted by Léna Gaubert
ceej @ceej.online · 16/07/2026
gonna start talking like this to my boss
screenshot where claude is saying "Good catch -- I fabricated that figure."
6180801214
Léna Gaubert @kittog.bsky.social · 13/07/2026
every day spent being surrounded by avid claude users feels like this (i'm being rage baited into the sun)
Group of men standing smiling, surrounded by the following texts from left to right: "let's see what Claude thinks", "I asked Claude", "Claude cooked here", "You should use Claude for that", "Did you ask Claude?", "Just use Claude", "Claude told me", "Claude's actually pretty good at this", "Let me throw that into Claude"
030
Reposted by Léna Gaubert
pengrui-han.bsky.social @pengrui-han.bsky.social · 30/06/2026
The human brain is strikingly modular: distinct networks for language, formal reasoning, social reasoning, physical reasoning. Is this fundamental to intelligent systems, or an accident of evolution? In our new preprint, we find the same modular organization emerges in LLMs.
129320
Reposted by Léna Gaubert
blair @blairaf.com · 02/06/2026
I'm looking for a new PhD student to come join me at UAlberta & Amii to study trans perspectives on AI policy & the future of AI as part of my new lab. The position is fully funded and offers access to a range of supports - more details in the poster below!
Flyer with qualifications and details for the PhD position.
37263
Reposted by Léna Gaubert
Les linguistes atterrées @tract-linguistes.org · 13/06/2026
« Le niveau des élèves en orthographe a baissé » : vrai ou faux ? Réponse courte : oui en gros… Réponse (plus) longue : c'est plus compliqué que ça… 🧵⬇️
Image introductive d'un reportage de M6 intitulé "Orthographe : le niveau des écoliers a baissé".
En arrière-plan, la pointe d'un stylo rouge est posée sur une copie d'examen.
43222
Léna Gaubert @kittog.bsky.social · 12/06/2026
Dublin & Howth on film! :) Shot on Pentax epsio w Mira 400 🍀
Street of DublinHowth harbourTemple Bar, DublinLast frame of my film, side looks burnt. View of Howth's peer.
121
Reposted by Léna Gaubert
Guillaume Plique @yomguithereal.bsky.social · 02/06/2026
xan v0.58.0 is out \o/ More performance, less cruft, very wow! You can now draw synthwave sparklines or visualize network layouts in your terminal like it's 1970! Full changelog: github.com/medialab/xan... @medialab-scpo.bsky.social
184
Léna Gaubert @kittog.bsky.social · 29/05/2026
turn off the mac book pro!!! enjoy the sweaty crowded streets and expensive pints of cider!!!
terminal console reads: it's not exactly what i wanted b-Canal St Martin, Paris, Summer evening. Quays are packed with people.
010
Léna Gaubert @kittog.bsky.social · 29/05/2026
maybe it's time we let the ai engineers go outside and have a break. maybe we can tell Claude "see you never". maybe we put the datacenters on pause too.
Text reads: is it okay if we meet and I prompt Claude to finish the task for some time?
100
Léna Gaubert @kittog.bsky.social · 25/05/2026
Dublin definitely got more "touristy", which means the city centre is absolutely crowded, just like any other big capital. you spend more time trying to avoid bumping into people than enjoying the city. Stephen's green is still as lovely as it used to be, we even got to sit in front of Joyce.
James Joyce's bust at Stephen's green, Dublin. 
Written on the stone: "James Joyce / 1882-1941 / Crossing Stephen's, that is my green..."
000
Léna Gaubert @kittog.bsky.social · 25/05/2026
felt very good to be back after so long (almost seven years, last time was 7 years ago). for some weird reason, we've had the best weather Ireland could ever have at this time of the year. it feels very wrong to get sunburned in Malahide.
View from living room.Asparagus being cooked in a pan.Kilrock Road, Howth
100
Léna Gaubert @kittog.bsky.social · 25/05/2026
some news from Dublin's bay. 🍀 surprisingly had too many conversations about the daily usage of Claude to organise meal prep and work weeks.
sun setting on Howth's harbour.at the back, Ireland's eye. view from Howth, co Dublinsail wrapped up. the words "tough nut" quilted on the wrap, white on orange.Max, laying down, next to the couch
110
Reposted by Léna Gaubert
Éditions Libertalia @libertalialivre.bsky.social · 22/05/2026
[Montreuil] Ce week-end le soleil sera au rendez-vous pour le 23e Festival des murs à pêches. Retrouvez Libertalia à la Prairie, 61 rue Pierre de Montreuil, samedi et dimanche, de midi à la nuit. Petit conseil: venez tôt, ce sera la foule. www.federationmursapeches.com
Affiche du festival des murs à pêches: beaucoup de vert (végétation) et un soleil incandescent.Visuel annonçant le stand de Libertalia aux Murs à pêchesPlan des parcelles des murs à pêches
053
Reposted by Léna Gaubert
Karen Hao @karenhao.bsky.social · 20/05/2026
On the one-year anniversary of EMPIRE OF AI, I am so, so excited to announce The AI Resist List, a new project that documents examples of resistance to the AI empires around the world 😍 airesistlist.org
The gorgeous AI Resist List homepage
4525291122
Léna Gaubert @kittog.bsky.social · 18/05/2026
Oooh that looks so cool!! Been a while since I did anything speech processing related, need to dive back into this someday. Excited to read this! (funny to see my name as an acronym tho haha)
010
Reposted by Léna Gaubert
arxiv cs.SI @arxiv-cs-si.bsky.social · 12/05/2026
An Experimental Method to Study Opinion Diffusion in Human-AI Hybrid Societies arxiv.org/abs/2605.09197
011
Léna Gaubert @kittog.bsky.social · 12/05/2026
We have many exciting ideas for this work, looking forward to discussing them in Rio this summer! 🇧🇷 6/6 🔗 Pre-print available on arXiv: arxiv.org/abs/2605.09197
arxiv.org
An Experimental Method to Study Opinion Diffusion in Human-AI Hybrid Societies
As artificial intelligence increasingly mediates public discourse, it becomes important to understand how human-AI collectives shape opinion formation, deliberation, and democratic outcomes. We presen...
000
Léna Gaubert @kittog.bsky.social · 12/05/2026
A biiiiig thank you to @norijacoby.bsky.social for your mentorship throughout this project! This definitely wouldn't have been possible without you. 🫶 And of course to my co-authors: Rémi Devaux and Antoine Jardin (Arlequin AI), Elif Çelen, Raja Marjieh and Diana Mangalagiu ☺️ 5/6
100
Léna Gaubert @kittog.bsky.social · 12/05/2026
As LLMs become more and more embedded in social media, we believe it is crucial to study their impact on collective opinion. Our LLM-only simulations (using Grok) suggest that prompt-framing might be a factor influencing polarization outcomes and large-scale opinion dynamics. 4/6
Evolution of Mean Polarization (top) and Mean Neighbor Correlation Index (bottom) over iterations in the Grok-only condition across the two instruction framings: opinion (green) and consensus (dark orange).
Shaded area represent +/−1 standard error of the mean across 21 runs per condition.
110
Léna Gaubert @kittog.bsky.social · 12/05/2026
Although very preliminary, our first (counterintuitive!) results suggest that hybrid social networks reach lower polarization faster than human-only networks. Human-only networks also happened to showcase lower neighbor agreement! 3/6
Evolution of Polarization
(top)and Neighbors Correlation Index (bottom) over iterations
across the three experimental conditions: human-only (pink),
Grok-only (dark orange) and hybrid (steel blue). Shaded area
(Grok-only) represents+/−1 standard error of the mean across
21 runs.
110
Léna Gaubert @kittog.bsky.social · 12/05/2026
How do AI-human interactions shape opinion dynamics? What role do LLMs play in online debate? To tackle these questions, we conducted online controlled experiments with a novel method we developed and ran three different setups: hybrid, human-only, and LLM-only networks. 2/6
Schema of our experimental design. A. Humans or AI agents were embedded in a 5 × 5 social network that evolved over eight
iterations. The network was initialized with positive opinions in the upper half and negative opinions in the lower half (supporting
vs. not supporting the central claim). In each trial, participants viewed three to five opinions from neighboring nodes in the
previous iteration (up, down, left, right, and the prior node at the same location) and selected the statement they felt best
represented the group. They were then asked to rephrase the statement to be consistent with prior observations. B. We tested
three conditions: human-only, AI-only (Grok), and hybrid (human–AI). C. An LLM was used to annotate each statement for
positivity, negativity, and neutrality.
120