Following the model release a few weeks ago, our collegues from the Falcon team released the technical report about Falcon-H1 training! Paper: arxiv.org/pdf/2507.22448 HF Model collection: huggingface.co/collections/ti…

👀 Time to do some reading :)

Contrastive divergences vs. contrastive learning (machine learning terminology can sometimes be confusing!)

And I was hoping it would know that a model for E[y|x] is regression.

Given the flattening we see in LLM performance, is this the right time to say that Yann LeCun was right, and transformers that trained on language are not "all you need"?

I have known Yam for a very long time, both personally and professionally. Everything he does is done with exceptional professionalism both in terms of product and engineering.

👀

🥳🥳 Happy to share that we have three papers accepted to EMNLP 2025 🇨🇳 (2 main, 1 findings)! What makes this special is that all three belong to a new research line I began last year: LLM-as-a-judge/LLM-as-an-annotator 🤖🧑‍⚖️

95% of Machine Learning solutions in the real world are for tabular data. Not LLMs, not agents, not fancy stuff. Learning feature engineering and building tree-based models will open a ton of opportunities.

Thrilled to announce TopoBench has been accepted to DMLR! 🚀 It’s a modular library for Topological Deep Learning, built to provide reproducible, cross-domain benchmarks and accelerate research. GitHub: github.com/geometric-inte… #AI #DeepLearning

AC cannot properly do their job if they don't read each paper in their batch, in addition to the reviews and author responses. Only then, they can find cases that the reviewer consensus needs to be overturned (to accept or reject) and can discuss those with the reviewers. SAC

This is simply an amazing idea!

AGI