Never thought that Deepseek and shawarma could be in the same sentence 😂😂

🚀 New Paper Alert! 🚀 We introduce Q-Filters, a training-free method for efficient KV Cache compression! It is compatible with FlashAttention and can compress along generation which is particularly useful for reasoning models ⚡ ⬇️R1-Distill-Llama-8B with 128 KV pairs ⬇️ 🧵

So happy someone (Marcel Salathé) did the work and put an end to the stupid take that ChatGPT etc. require a ton of energy 🙏🏼

⚡ Introducing truly Flash DeBERTa implementation ⚡ DeBERTa remains a top-performing model in Named Entity Recognition, Text Classification, and Reranking. Let's just make it more efficient, meet #FlashDeBERTa, which is up to 3-5 times faster. github.com/Knowledgator/F…

Need blazing-fast classifier inference with minimal code? ModernBERT now runs on vLLM — fast enough to process 200K+ arXiv papers in minutes. It makes running any of the 100s of ModernBERT models on the Hugging Face Hub even quicker. Guide 👇 danielvanstrien.xyz/posts/2025/vll…

I just read this WSJ article on why Europe's tech scene is so much smaller than the US's and China's. I'm afraid that, like most articles on this topic, it largely misses the mark. Which in itself illustrates a key reason why Europe is lagging behind: when you fail to

Arnaud Bertrand Europe has original ideas and the technical talents to implement them. The main reason why the European tech industry is small is a mistaken assumption of technological inferiority on the part of the European media. Perhaps more importantly, there was a similar inferiority

I think people not sure how crazy the climate change mindvirus is in Europe So they think I make up these stories But I don't lie, daily occurence, everything here is eco and about climate change and degrowth

Should we just focus our pre-training efforts on decoders? To answer this, we trained Ettin, various identically trained encoders and decoders, ranging from 17M to 1B parameters on 2T tokens of open data (beating Llama 3.2 and ModernBERT in the process)!

DeepMind founder Demis Hassabis on how he set up his life to maximize deep thinking with two full workdays in a 24-hour period: