🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

✨ Your know-how is your edge, AI is made to enhance it on your terms! 👀 Would you let AI access data you wouldn’t share with your team? 🛡️Internal governance remains paramount! This isn’t just smart permissions, it’s safer intelligence — activated 📩 Contact us to see how

In a time of abundant tech we began to not * bookmark sites but Google them over and over * compare photos by putting them in the same folder but screenshot and compare them by swiping left and right on a screen What are the new habits in world of abundant LLMs?

Among all those LLM releases, here is an important retrieval release: To overcome limitations of awesome ModernBERT-based dense models, today LightOn is releasing GTE-ModernColBERT, the very first state-of-the-art late-interaction (multi-vectors) model trained using PyLate🚀

Ah! I must have guessed ModernColBERT will have 8K context window. And I suspect that unlike 8K context windows from single-vector models (which notoriously rarely work), I'd expect this thing to actually just work. Incredible work Antoine Chaffin & LightOn team.

💫 LightOn releases GTE-ModernColBERT, the first State-of-the-Art late-interaction model trained on PyLate! By leveraging ModernBERT architecture and LightOn innovative PyLate library, this open-source multi-vector model sets a new milestone in the field and addresses the

In this example by Grant Sanderson Superposition is nothing more than compressive measurement or group testing but in a way that elates a deep sort of awe: it's quantum. We were so advanced.

It's funny, Grant even uses words like coherent. More interesting is the need to *not* define that function. The point I am probably making to myself is that for a long time I did not understand "Superposition" when in fact, it was a CS thing. Great job Grant Sanderson !

Quantum computing is a specific set of algorithms that try to do better than traditional or randomized algos "Try" is a key because noisy quantum hardware. My summary of all this ? bad noise = bad theory people.

Yes, I agree with Leon and said as such before. The impact is as large as writing.

I clearly remember someone saying that all the drugs were already known in 1970s but that the slow process of categorizing and establishing regimens of drug over time led to this large increase in survival over 30 to 40 years.

The "Edge of Tomorrow" plot is simply not believable. Bus line 96 simply never goes through the Place de la Concorde.

The new GTE-ModernColBERT-v1 by LightOn is the new #1 trending search/embedding model as of today. Definitely check it out if you're curious about multi-vector (a.k.a. late-interaction, ColBERT) search models.

ModernBERT 🤜 🤛 ColBERT = ModernColBERT ModernColBERT is the latest state-of-the-art text-based late interaction model by LightOn. - Based on the ModernBERT-base encoder - Outperforms ColBERT-small on the BEIR retrieval benchmark Here’s a quick notebook I threw together

Incroyable

Pourquoi est ce que la France subventionne une electricité moins chère pour l'Allemagne et l'Espagne? Deux pays qui ont montré un sous-investissement chronique dans leurs infrastructures et qui ont mis en danger tout le réseau européen?

As in LightOn 's logo!

Your Team deserves instant empowerment, for every new hire, from day one. 🚀 Unlock full access to your company’s know-how, instantly. 🙌 No IT tickets. No delays. Complete autonomy for newcomers. 🔐 Permissions stay synced across all your workspaces, always secure. ⚡ Automated

🎙️« L’enjeu pour les entreprises est d’arriver à extraire du sens à partir de n’importe quel type de contenu, et à fournir une réponse contextualisée instantanément.” Antoine Chaffin nous livre la vision de LightOn sur le RAG Multimodal (Retrieval Augmented Generation).

Cables are hard.

It's friday!