our new work on controlling recsys with natural language, led by @MicahCarroll, and with great collaborators Addie Foote, @kjfeng_ , Marcus Williams, @ancadianadragan, @wbradknox https://t.co/5mZdCDNZTP

What really matters in matrix-whitening optimizers (Shampoo/SOAP/PSGD/Muon)? We ran a careful comparison, dissecting each algorithm. Interestingly, we find that proper matrix-whitening can be seen as *two* transformations, and not all optimizers implement both. Blog:

Autoregressive language models learn to compress data by mapping sequences to high-dimensional representations and decoding one token at a time. The quality of compression, as defined by the ability to predict the next token given a prompt, progressively improves (as measured by

LLMs have shown a remarkable ability to “self-refine” and learn from their mistakes via in-context learning. But in robotics, most methods are single-shot. How can we bring inference-time adaptation to robot learning? A 🧵:

New work from @aditya_oberai & @seohong_park: instead of 1-step TD backups or n-step, can we "divide and conquer" over the trajectory, backing up finer and finer increments? Improves over bias of TD-0 and variance of MC. Principle is old, but getting it to work takes some care!

TD Learning can suffer on long tasks: ↑ deep bellman recursions → ↓ poor scalability (despite big data) We introduce a new method (TRL) with a "divide-and-conquer" value update, which scales well with long horizons!

🌍 LLMs can use long chain-of-thought (CoT) to reason in English, but what about other languages? New paper w/ @BerkeleyNLP: We study how scaling, pretraining, post-training & inference affect long CoT across 9 languages. Spoiler: English long CoT ≠ multilingual long CoT 🧵

Our new paper with Sonali Sharma and @RoxanaDaneshjou is out in @npjDigitalMed! We examine how medical safety and disclaimer messages in public LLMs have changed over time when answering patient questions.

AI can now see, reason, and segment the Earth. 🌍 Meet LISAt, our #NeurIPS2025 Datasets & Benchmarks paper - the first foundation model that turns language queries into pixel-level satellite segmentations. 🛰️ (1/n) 🔗 https://t.co/ApVZgGF0cU @NeurIPSConf @berkeley_ai

Can a robot inspect all views of an object? Today @IROS, we present Omni-Scan from @berkeley_ai, a novel method for bimanual robo 360° object scanning & reconstruction using 3D Gaussian Splats. (1/8) 🔗 https://t.co/8emyJfUNk4

🧠 New preprint: How Do LLMs Use Their Depth? We uncover a “Guess-then-Refine” mechanism across layers - early layers predict high-frequency tokens as guesses; later layers refine them as context builds Paper - https://t.co/5PitHjmJJZ @neuranna @GopalaSpeech @berkeley_ai

Catch up on our most recent Community Lecture: “Transmission Versus Truth: What Will It Take to Make an AI as Smart as a 4-Year-Old?” with Alison Gopnik.  This was the last of six community lectures for 2025, and all are available to watch on SFI’s YouTube channel. Watch here:

New evaluation results from @AnthropicAI's Claude Sonnet 4.5’s system card on our CyberGym benchmark reveals a striking trend: AI cybersecurity capabilities are advancing at unprecedented speed—from ~10% (Claude-Sonnet -3.7) to ~30% success rates (Claude-Sonnet-4.5) (with single

Amazing! 10 @BerkeleyEECS @SkyCompLab grad students are Amazon AI PhD Fellows! Congrats! Learn more about our fellows here: https://t.co/zuCGKlmSNe #AmazonAIFellowship @BerkeleySky

Humans handle dynamic situations easily, what about models? Turns out, they break in three distinct ways: ⛔ Force Stop → Reasoning leakage (won’t stop) ⚡️ Speedup → Panic (rushed answers) ❓ Info Updates → Self-doubt (reject updates) 👉Check out https://t.co/wKrnsMkiFY

Simulation drives robotics progress, but how do we close the reality gap? Introducing GaussGym: an open-source framework for learning locomotion from pixels with ultra-fast parallelized photorealistic rendering across >4,000 iPhone, GrandTour, ARKit, and Veo scenes! Thread 🧵

How can a robot provide details of plant anatomy for plant phenotyping? Today @IROS2025 , we present Botany-Bot from @berkeley_ai @Siemens. Botany-Bot 1) creates segmented 3D models of plants using Gaussian splats and GarField 2) uses a robot arm to expose hidden details. (1/9)

✨Introducing ECHO, the newest in-the-wild image generation benchmark! You’ve seen new image models and new use cases discussed on social media, but old benchmarks don’t test them! We distilled this qualitative discussion into a structured benchmark. 🔗 https://t.co/wJmmEY8TFQ

Do you have many models to choose from and little labeled data with which to evaluate them? Check out our #neurips2025 paper, which presents a method to estimate model performance more accurately than previous methods using both labeled + unlabeled data.

Super excited about @wenjie_ma's work on verifying math proofs! ✅ 24 competitions, 3 SoTAs (o3, Gemini-2.5-Pro, R1) ✅ Strong evaluator -- a carefully designed evaluator with simple ensemble beats agentic ones ✅ Strong best-of-n performance Check out the paper & website!