🎯 Learning is most effective at the frontier of capability: problems that are too easy or too hard teach nothing. For LLM reasoners trained with GRPO this is literal: problems the model always or never solves give zero gradient. We introduce Frontier Learning👇🧵

arXiv has updated our policy on rate limiting for all submitters. This update was made to fairly distribute moderator time & support the arXiv community of staff, volunteers, readers & authors. Please read our announcement to learn more: blog.arxiv.org/2026/10/01/up…

🚨 We turned European train delays into a live league table. Germany finished last. 🇩🇪 Switzerland isn't even #1. 🇨🇭 Over the last 30 days, DelayBahn tracked 900,000+ train stops across 6 countries. Welcome to Europe's Rail Leaderboard 🏆🚆 See where your country ranks 👇 delaybahn.com/rangliste Some of the results are wild 🧵 (1/6)

#AISTATS 2027 Call for Paper is out! This year's edition is taking place in Montreal on May 3-6th, 2027 🇨🇦 Aymeric Dieuleveut and I are chairing virtual.aistats.org/Conferen… New this year: AI review, and more! Abstract deadline: September 29, 2026 Paper deadline: October 6, 2026

Important points. Note also that users can opt out from use of their (de-identified) data in training. Anthropic has similar policies, though I do not remember such discussions when they announced mathematical results.

Anthropic: “Claude Team plan for Scientists!” Me: applies Anthropic: “We were unable to verify that you are a scientist.”

London’s Hyde Park has transformed into the Dubai desert.

Nice. Didn't know this was an option.

Releasing the model weights and technical report of Kimi K3. Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window. New model architecture: 2.5x the intelligence per unit of compute, not just more params. Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale. Model weights: huggingface.co/moonshotai/Ki… Tech report: github.com/MoonshotAI/Kimi-K… Tech blog: kimi.com/blog/kimi-k3

Slides from my ICML tutorial "Is numerical optimization theory irrelevant to machine learning practice in 2026?": cs.ubc.ca/~schmidtm/Document… (Updated to fix some typos, incorporate feedback, and add some things I did not have time for. Will link to the video here when posted.)

(1/6) A Warp Drive for Neural PDE Solvers??? Tomorrow in Seoul, Till and Alex are presenting their floral #ICML2026 Spotlight. Don’t miss it if you care about neural surrogates for physics, foundation models, physics world models, …

Judging optimizer gaps by looking only at language modeling with a fixed batch size is dangerous: one gets only 1/2 of the story. @orientino_ and @ruuustem_10 went beyond. Turns out that for every model, task, and data, there is always a setup where Adam > SGD. 🧵

Hiring 2 summer ML research interns at the University of Basel 🇨🇭. Research topics: RL/diffusion LLM post-training, reasoning, or LLM orchestration. Possible fully funded PhD offers to follow. I'll be at ICLR this week and happy to chat. Apply: forms.gle/TeeeNU6e7kDH3jX96

After a lot of hands-on time with Codex and Claude: Claude is significantly better at actually following instructions. Not even close right now.