🤔 How do we train AI models that surpass their teachers? 🚨 In #COLM2025: ✨Delta learning ✨makes LLM post-training cheap and easy – with only weak data, we beat open 8B SOTA 🤯 The secret? Learn from the *differences* in weak data pairs! 📜 arxiv.org/abs/2507.06187 🧵 below

Life itself is meaningful enough...

Tokenization has been the final barrier to truly end-to-end language models. We developed the H-Net: a hierarchical network that replaces tokenization with a dynamic chunking process directly inside the model, automatically discovering and operating over meaningful units of data

Change always comes later than we think it should.

It's not solely the culpability of who has more; it's also the culpability of those who want more, more so than the former.

Without downplaying the achievements of the language models, a level headed view of looking through the hype. Terence Tao on the IMO gold medal wins of Gemini/OpenAI

This is pretty amazing

🤯

Hope is realizing that 99.9% of all species that ever lived are extinct, and if we keep pushing, the Earth could shake us off like a bad cold in the blink of an eye.

🌱 Amazing documentary about our world, nature is more alive than we think. Some surprising facts I learned: - Beyond sensitivity to temperature, sunshine, and humidity, plants can also smell, touch, taste, hear, and perceive shapes. - Plants have a vascular system similar to our

Today we are announcing Genie 3, a general purpose world model by Google DeepMind that can generate dynamic, interactive environments with a single text prompt. World models are AI that understand facets of the world (like Veo's knowledge of intuitive physics or Genie's mastery

I want my WiFi in my forest

Perhaps the reason we can't explain what mass or charge truly is, is because we're interacting only with the user interface of reality. Evolution may have abstracted away the underlying mechanisms of how the universe actually works.

I remember being excited about AI. I remember 20 years ago, being excited about neuroevolutionary methods for learning adaptive behaviors in video games. And I remember three years ago, mouth watering at the thought of tasty experiments in putting language models inside

x.com/i/article/1963…

The map may look wrong until the terrain shows itself.

.Graham Neubig and I are co-teaching a new class on LM inference this fall! We designed this class to give a broad view on the space, from more classical decoding algorithms to recent methods for LLMs, plus a wide range of efficiency-focused work. website: phontron.com/class/lminfere…

Why do we need GPUs for AI? ➡️A GPU (Graphics Processing Unit) is built for parallelism – it splits a bigger job into smaller tasks and distributes them across processing cores. Inside a GPU, billions of tiny transistors are etched onto a silicon chip, arranged into thousands