
From our team at @GoogleDeepMind: we ask, as an LLM continues to learn, how do new facts pollute existing knowledge? (and can we control it) We investigate such hallucinations in our new paper, to be presented as Spotlight at @iclr_conf next week.

RSI @GoogleDeepMind ex-IMO (🇨🇦) ex-neuroscientist Views are my own

From our team at @GoogleDeepMind: we ask, as an LLM continues to learn, how do new facts pollute existing knowledge? (and can we control it) We investigate such hallucinations in our new paper, to be presented as Spotlight at @iclr_conf next week.

Check out our new model. Awesome for automated scientific discovery!
Gemini 4 Carbon is now available. It’s our strongest model, built to dig deeper, handle long-horizon tasks, and multi-step reasoning. If Argon is great for complex tasks, Carbon is the model you bring in for your hardest problems.

Today we are happy to share one of our first open research releases: Latent Sparse Attention (LSA), a step towards subquadratic attention that matches accuracy of full attention at ultra-long context. Sparse attention attends to only a few tokens per query, but it has to find the right ones. At 10M tokens, that search takes 98% of the attention layer's time. LSA makes sparse attention up to 22.7× faster and full attention 72.4x faster. With @kubicek_ales @caciolai @PengLiangzu @SukritSumant @Ben_Hoov @FraserGreenlee @RMBattleday @radim_urban @thoefler @jcrwhittington

Denario is out in PRX Intelligence: a modular multi-agent research assistant that can go from idea → literature → methods → code & plots → paper draft across disciplines — built with AG2 / Cmbagent. 📄 journals.aps.org/prxintellig… arXiv: arxiv.org/abs/2510.26887

Fake news.
$GOOGL EMPLOYEES RAISE CONCERNS OVER GEMINI 4 CODING PERFORMANCE Google is preparing to launch Gemini 4, but some employees with access to the model say its strong benchmark results aren’t fully translating into real-world performance, according to Bloomberg. The main concerns are around coding. Some internal testers say Gemini 4 struggles with certain coding tasks and front-end design, while the model is also said to be very large, potentially making it more expensive to run. Google disputes that Gemini 4 is underperforming and says there is broad internal consensus that the model is at the frontier. Bloomberg also reports Google abandoned Gemini 3.5 Pro after previously planning to release it in June. Internally, views appear mixed: some employees believe Gemini 4 still trails the latest models from OpenAI and Anthropic in certain areas, while others believe it has caught up. Gemini 4 is said to perform particularly well on multimodal tasks, including understanding video, along with safety, cybersecurity and natural communication. Source: Bloomberg

Can we build world models purely in the latent space, without pixel prediction? We present Contrastive World Models - we train the latent states to maximize mutual information with future observations, without any decoders / pixel reconstruction. Contrastive World Models enables ⚡ Substantially more robust representations 🤖 More efficient training by removing the pixel decoder entirely 🌍 A general approach with minimal assumptions [1/n]

1/ Today, we introduce Faraday, a 27B-parameter AI Scientist that extends the capabilities of coding agents with a layer of scientific intuition. Trained via long-horizon RL, Faraday outperforms Claude Opus 4.8 and GPT-5.5 on the task of replicating research papers. 🧵

We’re excited to announce DiG-bench, a new benchmark for discovery! Over the last few weeks we’ve been testing frontier AI models on our novel discovery games and seeing how they score. Each game is a text-based environment, so they probe discovery capabilities in the natural domain of language models, rather than requiring additional, potentially confounding, visual understanding. TL;DR frontier models have improved a lot over the last few months. But they are still stumped by some surprisingly simple problems, even in their native text domain. With @cocosci_lab (@Princeton) @MITCoCoSci (@MIT) @SchmidhuberAI (@KAUST_News) @misovalko (@Inria) @tri_dao (@PrincetonCS) @RMBattleday @zebkDotCom @FraserGreenlee @akaijsa @ClareMaguire @TimMuller1 @kubicek_ales @physicscat0x7d @SukritSumant @thoughtchannel_ (1/5)

Working on agents? Submit your work to our workshop! Harness, meta-optimization, self-improvement, safety… you name it! We’ll have paper awards and free registration for students 🎫🎟️ let’s manage agents together 💪
🚀Call for Papers — @NeurIPSConf 2026 Workshop Workshop on Responsible Use of Meta-Agents 📅 December 11/12 · 📍 Sydney, Australia Join us to shape the future of Meta-Agents and Harness. Topics include but are not limited to: 1. Automated Design of Agent Harnesses 2. Optimization of Agentic Systems 3. Recursive Self-Improvement 4. Evaluation and Benchmarks for Meta-Agents 5. Misalignment and Safety for Meta-Agents 6. Governance & Human Oversight of Agents 🌟 Featuring an all-star speaker lineup: Graham Neubig @gneubig (CMU) Chelsea Finn @chelseabfinn (Stanford) Jenny Zhang @jennyzhangzt (Recursive) Robert Lange @RobertTLange (Sakana AI) Bo Li @uiuc_aisecure (UIUC) Hancheng Cao @CaoHancheng (Emory) Yuandong Tian @yuandongtian (Recursive) Chen Sun @ChenSun92 (Google DeepMind) Proudly sponsor by @metaai and @e2b Updates: Follow us here & spread the word! #neurips2026 #metaagents #llms #harness #rsi #selfimprovement

Looking forward to being a panelist on this! It’s going to be a blast💥
🚀Call for Papers — @NeurIPSConf 2026 Workshop Workshop on Responsible Use of Meta-Agents 📅 December 11/12 · 📍 Sydney, Australia Join us to shape the future of Meta-Agents and Harness. Topics include but are not limited to: 1. Automated Design of Agent Harnesses 2. Optimization of Agentic Systems 3. Recursive Self-Improvement 4. Evaluation and Benchmarks for Meta-Agents 5. Misalignment and Safety for Meta-Agents 6. Governance & Human Oversight of Agents 🌟 Featuring an all-star speaker lineup: Graham Neubig @gneubig (CMU) Chelsea Finn @chelseabfinn (Stanford) Jenny Zhang @jennyzhangzt (Recursive) Robert Lange @RobertTLange (Sakana AI) Bo Li @uiuc_aisecure (UIUC) Hancheng Cao @CaoHancheng (Emory) Yuandong Tian @yuandongtian (Recursive) Chen Sun @ChenSun92 (Google DeepMind) Proudly sponsor by @metaai and @e2b Updates: Follow us here & spread the word! #neurips2026 #metaagents #llms #harness #rsi #selfimprovement

Little known fact: Jacob Tsimerman was a beast from a young age. IMO gold 🏅, and a rare perfect score. A few years later he coached us on Team Canada. One of the kindest guys I've ever met.
Mathematician Jacob Tsimerman from @UofT has won the 2026 Fields Medal. Tsimerman extended the scope of o-minimal techniques within arithmetic and complex algebraic geometry and proved Griffiths' conjecture on the algebraicity of images of the period maps. He is also a Simons Investigator in Mathematics. Watch his full video and see all of the Fields Medal winners: bit.ly/4psOJbU #math #icm2026 #fieldsmedal

@ClementBonnet16 and I started After Labs with a simple conviction. The next leap in AI will not come from models that merely store more knowledge or bake more task-specific skills into their weights. It will come from systems that can continually learn and refine their understanding of the world through interaction. At After Labs, we are singularly focused on developing world models with the ability to adapt with every new data point instead of staying static. These adaptation mechanisms need to be built into the model during training and not bolted on as an afterthought. Stay tuned as we pursue this mission of building curious agents that never stop learning 💡
Current models rely on brute-force RL training to adapt to new tasks and environments. However, intelligent systems should be able to learn on the fly in a sample-efficient manner, continuously refining their understanding of the world through their interactions. After Labs is developing models with built-in adaptation mechanisms rather than treating adaptation as an afterthought. This new generation of models will start a new era of curious agents that may not possess all the world's knowledge but will have the crucial ability to learn in real time. Today, we are happy to announce that After Labs has been selected as one of 10 AI labs to join NFAI and receive €3M in additional funding. This support will accelerate our mission to build models that continuously learn about the world. We will be contributing to open science along the way, so stay tuned and follow us on our journey to develop AI that augments human understanding of the world 🌍

We lost our beloved PhD mentor, Susumu Tonegawa, leaving an emptiness. Whenever I think about courage, I think about Susumu. He switched fields twice and relearned everything. Whenever I think about greatness, I think about Susumu. He was on a mission to discover how Nature works for the benefit of humanity. Whenever I think about perseverence, I think about Susumu. There is a funny anecdote we knew that when he climbed the great wall of China at a scientific conference, he wanted to race everyone to the top and was perplexed when all the (younger) faculty members were much slower than he was. He was a lion of a man. A dear friend from the lab once told me that seeing science through his eyes has been one of the highlights of his life. It is mine as well. I think this poem that by Ki no Tsurayuki expresses well what I feel , which I think others who have lost mentors can also feel 色も香も昔の濃さに匂へども 植ゑけむ人の影ぞ恋しき Blossoms deep in color and fragrance as in the old days, yet I long for the one who first planted them. Photos from happier days:

ぽっかりとこころに穴が空いた朝。師であり、父でした。Susumu先生がいなければ、今の自分はいなかったと断言できます。きっと向こうに行っても、いつものように「Engramがー!」とパワフルに叫びながら、Xu達とディスカッションしているのでしょう。本当にただただ悲しい。安らかにお眠りください。

We're building a team at @Anthropic focusing on AI and the rule of law. We've made our first hires, and are now opening up a new research engineer role. We're looking for people with advanced technical skills, including AI/deep learning/NLP, full-stack development and data science, paired with training or experience in law, government, political science, or a related field. If this is you, or a friend, please get in touch. job-boards.greenhouse.io/ant…

Everybody is talking about recursive self-improvement (RSI) and meta learning. Here is my old 2020 talk about this [1]. It has aged well. Example: humans still define the starts & ends of trials of many modern meta learners. My RSI systems since 1994 LEARN to (re)define them [2]! [1] Meta Learning Machines in a Single Lifelong Trial (talk for workshops at ICML 2020 and NeurIPS 2021, based on earlier talks since 1994). Abstract: the most widely used machine learning algorithms were designed by humans and thus are hindered by our cognitive biases and limitations. Can we also construct meta learning algorithms that can learn better learning algorithms so that our self-improving AIs have no limits other than those inherited from computability and physics? This question has been a main driver of my research since I wrote a thesis on it in 1987 [2]. Here I summarize our work on meta reinforcement learning with self-modifying policies in a single lifelong trial (since 1994), and mathematically optimal meta-learning through the self-referential Gödel Machine (since 2003). Many additional publications on meta-learning since 1987 can be found in the RSI overview [2]. [2] J. Schmidhuber (AI Blog, 2020-2025). 1/3 century anniversary of first publication on recursive self-improvement (RSI) and meta learning machines that learn to learn (1987). For its cover I drew a robot that bootstraps itself. 1992-: gradient descent-based neural meta learning. 1994-: meta reinforcement learning with self-modifying policies. 1997: meta RL plus artificial curiosity and intrinsic motivation. 2002-: asymptotically optimal meta learning for curriculum learning. 2003-: mathematically optimal Gödel Machine. 2020-: new stuff!