Amazing team from Astera Institute at Seeed Studio NVIDIA Robotics LeRobot home robot hackathon! ✨ VR app with stream & arm mapping 🎮 Isaac Sim BEHAVIOR env + RL 🗺️ MUSt3R-based navi & 3D reconstr 🧠 GR00T 1.5 training with Jetson Thor All open source:

Check out new work on Training LLMs to Discover Abstractions for Solving Reasoning Problems: Paper+Code: cohenqu.github.io/rlad.github.io/ Reasoning goes beyond pattern matching or memorization, it requires discovering and applying algorithmic procedures that reuse primitives,

New work on Rethinking Thinking Tokens: LLMs as Improvement Operators: arxiv.org/abs/2510.01123 Reasoning training encourages LLMs to produce long chains of thought (CoT), improving accuracy via self-checking but increasing context length, compute cost, and latency. This work

🚀 New Paper: Bottlenecked Transformers – Memory Consolidation for Better Reasoning in LLMs 🧠📚 Large Language Models have shown that more inference-time compute often means better reasoning. But most of this extra computation happens in token space — through longer chains of

What if robots could improve themselves by learning from their own failures in the real-world? Introducing 𝗣𝗟𝗗 (𝗣𝗿𝗼𝗯𝗲, 𝗟𝗲𝗮𝗿𝗻, 𝗗𝗶𝘀𝘁𝗶𝗹𝗹) — a recipe that enables Vision-Language-Action (VLA) models to self-improve for high-precision manipulation tasks. PLD

Aight let's unclickbait the fp16 paper. tł;dr cool paper, a little bit overstated in comms, very overstated by poasters. The thing that gave me a pause is that on the surface, it seems to claim that bf16 is horrible, borderline unusable. But that's not really the case (nor is

Here is the full run of MicroFactory, that autonomously assembles a photo frame - a real product sold on Amazon. This is a good example of how to deal with complex tasks, while we do not yet have a large set of robotic data to train a big general ai model. Thread:

The VCL paper has arguably the first example of modern continual learning for GenAI: VAEs trained on digit/alphabet images 1-by-1 arxiv.org/abs/1710.10628 Coded by yours truly ☺️ who was (and still is) 🥰 in generative models. Time to get back to continual learning again?

So much love and joint work in this OpenAI team 😂😂😂 #openai

This deserves much more attention.

If you have ever worked with ROS, then you know the pain! Ark elevates this pain by offering you a low-weight Python package that remedies that pain. 1⃣ pip installable ✌ 2⃣ Integrates many robots and sensors ✌ 3⃣ Supports a publisher-subscriber architecture 😎 4⃣ Offers

This is a new small LM (1.5B) that achieves impressing reasoning capabilities on AIME, LCB, and GPQA. The authors used model merging and an entropy maximizing variations of GRPO. Impressive work! 🔗arxiv.org/abs/2511.06221

Very interesting results from Cyrus (Cyrus Wai-Chung Kwan) -- training on generated math reasoning problems within an open-ended self-play framework can yield more accurate results than training on "gold" datasets like GSM8K or MATH!

I agree with Jeremy Howard

Chris Murphy 🟧 You're being played by people who want regulatory capture. They are scaring everyone with dubious studies so that open source models are regulated out of existence.

The AIRR-ML-2025 challenge is LIVE on @Kaggle! We’re challenging the best in #MachineLearning & #DataScience to decode the immune system. 💰 $10,000 prize pool 📄 Co-author a @NatureMethods paper

Do you want to see if your robot can learn a thousand tasks in a day ... ? Following our Science Robotics paper last week, we now give you: >> Our Code and Model Weights 🧑‍💻 >> An Explainer Video for the Paper 📽️ For both: robot-learning.uk/learning-1000-… Let us know how you get on!

I have not complained for a while, so I have this itch to do that 😆 They tell you robotics is almost there! They tell you it is just on the cusp! They tell you to get a robot to your house (and let a tele-operation engineer invade your privacy)! Well, well, well ...

hey #England, how did we go from this to this? what the heck happened! #uk Keir Starmer

chatted with a journalist earlier this week about temporal reasoning and VLMs, and one explanation that popped up for some behaviour is that those models are like a person waking up after a nap that lasted a few years -- this is very on point