Very cool! Hi Adina, do you know by chance if the 🤗 Paris office allows visitors? I’d love to buy some merch like a 🤗 t-shirt for example. 😅🤭

Rouen snapshots 🇫🇷🏰🧙‍♀️.

Disappointing that this was not the golden ratio…

TGIF.

Our DeBias-CLIP paper has been selected as a CVPR 2026 Highlight 🤩🤩🤩!

Obsession with symmetry and geometry back at it.

Immaculate geometry. 📐

Paris spring days hit different.

Finishing the evening jogging and getting some fresh air on top of the butte Montmartre 🙂.

Our DeBias-CLIP paper has been selected as a CVPR 2026 Highlight 🤩🤩🤩!

A bit of dreamy landscapes from Saint-Malo 🇫🇷😍.

I did something cool today. 🌊

The houses are pretty cute too.

Spending Easter Sunday in Rennes 🇫🇷👌🏻.

Somehow managed to get Céline Dion, the French flag and the Eiffel Tower in the same picture. 👌🏻🇫🇷🇨🇦

La vie nocturne de Paris. 🌚⭐️🌙.

Cool exposition on the work of Mickalene Thomas at Le Grand Palais, exploring themes of femininity, race, sexuality, and gender. 🖼️

Better weather today 🙂.

Also, my Parisian apartment (17e arrondissement) is the cutest thing ever 😍.

Happy to be back in Paris for a while. La météo n’est pas idéale, mais le printemps est tout de même là 🌸🌼🌺. 🙂

Really? Usually he prefers to grab other stuff 🙃.

And the codebase and open weights are now up and running!

Our model DeBias-CLIP (CVPR 2026) is currently the new SOTA (by a large margin) for long text retrieval!

Try it out if you’re doing image <-> text retrieval (text sequence up to 248 tokens). 🙂

I’m so excited I’m just gonna start book #2 from you (Reasoning model from scratch)! 😄😄😄

And don’t get me started on « towards » 🤣😅. How much towards? Just facing at the right direction, or taking one step, or rather sprinting towards it? 🙃🙃🙃

Or the gap is 10 meters long and their specific paper only fills up 20 cm of it 🙃.

We've been investing heavily in better protein language models (PLMs), but relatively little work addresses how to best generate with them. We present a new search-based method for PLMs and exhaustively benchmark models and methods, including with in vitro data from antibody therapeutics campaigns.🧵

👀

These are just clickbait posts to create engagement and get new followers. Also probably AI-generated. The author has no clue and has likely not even read the paper 😫.

After a series of fatal strikes involving civilians in Iraq and Afghanistan, Congress directed the Pentagon to reduce civilian casualties as part of a 2019 law. During the Biden administration, the Defense Department created the Civilian Harm Mitigation and Response initiative. The civilian mitigation teams – cut by 90% by Hegseth – work with military commanders on target planning, and making sure that targets are actually military sites. The teams help come up with "no strike" lists, including religious and cultural sites and schools

I knew that the Department of War shot a missile into a school & killed 175 civilians, mostly children.

This is learned today:

Secretary of War Hegseth had cut the teams assigned to prevent such tragedies by 90%.
www.npr.org/2026/03/11/n...

Also, notice how it's always quantum or crypto bros?