Google Research

Today we share an in-situ comparative research study of a set of experimental prototype AI agents exploring how conversational AI might conduct end-to-end symptom interviews & differential diagnostic assessment for research benchmarking purposes. More →goo.gle/3RQqt79

Today we announce a new paradigm for quantum control. By integrating reinforcement learning with quantum error correction, we enabled a quantum computer to continuously adapt to drift, stabilizing the system during computation. This improved logical stability 3.5x. Learn more: t.co/U1jTSmFrdQ

Vídeo

What drives a diffusion model's creativity (i.e., its ability to generate novel data rather than memorize)? Today we show this is a mathematical consequence of neural network regularization causing score smoothing and interpolation. Read the blog to learn more: t.co/menk2a3Xy6

GIF

No one should ever be surprised by a natural disaster. ✨ That is our north star for crisis resilience, as highlighted by @ymatias, GM of Google Research and Google Crisis Resilience Lead, where he also shares how we're using AI to predict floods, wildfires, and extreme weather to help keep communities safe globally. 🌍👇 Explore our work: t.co/Oy9FOWUap0 Read the latest blog on our progress: t.co/q4a8AKIOmQ

Vídeo

How does quantum computing help determine complex molecular structures? Google Quantum AI's Willow chip demonstrates the first-ever algorithm to achieve verifiable quantum advantage on hardware. Researchers from the team take you behind the scenes of this breakthrough that marks a significant step toward the first real-world applications of quantum computing. t.co/qEOuy0pHhv

In 2023, we collaborated with the World Health Organization to launch Open Health Stack, a suite of open-source tools to help developers build secure, next-gen digital health solutions. Today, we’re announcing that we’re transferring the code and assets to the Linux Foundation which will launch the new Open Health Stack Software Foundation (OHS-SF). This will provide a vendor-neutral, community-led home for the future of these essential building blocks for digital health. We’re excited to see how this empowers the developer ecosystem and local innovation. Learn more: t.co/r0q0MVJSRU

Introducing SensorFM, a large-scale Sensor Foundation Model that learns from 1 trillion-minutes of unlabeled wearable data drawn from five million consented participants. SensorFM learns a single, reusable representation of sensed human physiology that transfers across cardiovascular, metabolic, sleep, and mental health, as well as lifestyle and demographic factors. More →t.co/lbi1DG0zAW

Krzysztof Choromanski returns to the Google booth (#B206) to explore practical robotics research today at 3pm. Learn about Gemini Robotics models, 3D Transformers, and efficient techniques for object detection. #ICML2026 x.com/GoogleResearch…

Learn more about how Isomorphic Labs' Drug Design Engine is using artificial intelligence to help solve disease. Join Agnieszka Podsiadlo and Andreas Loukas today at 3:00pm at the Google booth (#B206) to discuss AI healthcare pipelines. #ICML2026

Join Vahab Mirrokni and Michael Galkin at 12:30pm today at the Google booth (#B206) for a session on graph foundation models. Learn how generalizable graph learning improves systems and coding, driving 20% datacenter efficiency gains and optimizing database queries. #ICML2026

How do we ensure multimodal foundation models remain reliable when faced with real-world distribution shifts? Meet Naman Goyal and Jenny Ni today at the Google booth (#B206) to explore how small changes like cropping, dimming, or compression can impact both a model's answers and its verification evidence. Discover why deployment reliability is tested trust, not just a clean benchmark score. #ICML2026

How can we map deep research ideas into practical robotics? Join Krzysztof Choromanski today at 11:00am at the Google booth (#B206) to explore Gemini Robotics models, 3D Transformers, and object detection techniques. #ICML2026

Pluralis v0.1 is a novel multimodal, multi-regional, and multilingual dataset built from a culture-first perspective. Meet Lora Aroyo today at 9:30am at the Google booth (#B206) to explore localized safety evaluation paradigms. #ICML2026

Congratulations to the @GoogleDeepMind authors of "Asynchronous Methods for Deep Reinforcement Learning", recipient of the #ICML2026 Test of Time Award. This work shows that asynchronous actor-critic succeeds on a wide variety of continuous motor control problems as well as on a new task of navigating random 3D mazes using a visual input.

Next up at the booth today at 3:00pm, explore how Gemini Omni unlocks video creation in new ways. Visit Sarah Xu, Miaosen Wang, and Minkai Xu at the Google booth (#B206) to learn more and view our live capabilities overview. #ICML2026

Co-Director is a hierarchical multi-agent framework that solves the challenge of narrative drift in long-video generation. Meet Yale Song and Yiwen Song today at 3:00pm at the Google booth (#B206) to see generative storytelling in action. #ICML2026 Watch more here! t.co/cCjcCU3Vel

Toby Staines and Michal Kazmierski are heading back to kiosk 1 at the Google booth (#B206) today at 12:30pm. Stop by to explore AlphaEarth Foundations and see how this embedding field model enables planetary geospatial insights. #ICML2026 x.com/GoogleResearch…

Next up at the booth today at 12:30pm, see how Gemini collaborates with artists to design complex origami. Join Tom Zahavy, Xidong Feng, and Shaobo Hou at the Google booth (#B206) to see abstract ideas turn into mathematically valid physical art. @GoogleDeepMind #ICML2026 Read more about it here: t.co/mUG1Wz7fCo

Next up at the booth today at 12:30pm, see how Gemini collaborates with artists to design complex origami. Join Tom Zahavy, Xidong Feng, and Shaobo Hou at the Google booth (#B206) to see abstract ideas turn into mathematically valid physical art. #ICML2026

Coral NPU provides a glimpse of the future of LLMs on low-power platforms. See English-to-target language translation running on a low-power Torq-Coral NPU with Gregory Kielian today at 12:30pm at the Google booth (#B206). #ICML2026

Meet Toby Staines & Michal Kazmierski today at 11:30am at the Google booth (#B206) to learn about AlphaEarth Foundations. This embedding field model yields a unified representation of the Earth's surface by assimilating multiple remote sensing data sources. #ICML2026

Visit the Kaggle Game Arena today at 9:30am at kiosk #2 where frontier models compete in live environments. Stop by to browse benchmarks and leaderboards with Yuting Han, Lucy He, and Tiffany Xiang at the Google booth (#B206). #ICML2026

How can we address the scarcity of data required for specialized AI? Join Hamza Harkous, Tim Davidson, and Cesar Ilharco today at 9:30am at the Google booth (#B206) to learn about Simula, a framework for synthetic dataset design. #ICML2026

GIF

Three new FireSat satellites are officially in orbit. 🚀 This launch is a major scaling milestone for the Earth Fire Alliance’s mission to provide continuous, high-resolution wildfire detection. See how we're working together with the @EarthFireAll & other partners to put AI to work for climate resilience: t.co/VOjbEvySIh

Today on the blog, read all about how we demonstrate the effect of coordination in routing on improving road traffic efficiency. We show that rerouting even a small fraction of trips to disperse traffic can measurably improve driving speeds. More →goo.gle/4p5BL3A

GIF

Experience AI Scientist, a novel multi-agent system designed to automate the end-to-end scientific research pipeline. Meet Jinsung Yoon and Rui Meng today at 6:00pm at the Google booth (#B206) to learn how complex workflows reliably accelerate AI discovery. #ICML2026

Next up at the booth today at 3:00pm, explore a general sound separation system based on the fully unsupervised training of a generative flow matching model. Meet Henry Li and Robin Scheibler from @GoogleDeepMind at the Google booth (#B206) to listen to the output. #ICML2026

Can video models become general-purpose foundation models for vision? Join Priyank Jaini today at 3:00pm at the Google booth (#B206) to see how Veo 3 solves a broad range of tasks without explicit training, showcasing early forms of visual reasoning. #ICML2026

Yana Lunts, Sam Holt, Pavel Dubov, and Stefan Moser will be back at kiosk 2 at the Google booth (#B206) today at 3:00pm. Drop by to experience Project Astra firsthand and meet the team behind the research prototype. #ICML2026 x.com/GoogleResearch…

Curious how the universal assistant of tomorrow might look? Learn about Project Astra, a research prototype of a next-generation AI video and audio assistant. Meet Yana Lunts, Sam Holt, Pavel Dubov, and Stefan Moser today at 1:00pm at the Google booth (#B206). #ICML2026

Explore Vision Banana with Shangbang Long to experience state-of-the-art computer vision tasks including segmentation and depth estimation built from Nano Banana. Join us at the Google booth (#B206) today at 12:30pm. #ICML2026

How do we choose what models to run and what input to give them for every query that goes to Google Search? Join Andrew Gilchrist-Scott, Manik Panwar, Zihan Jiao, and Ramon Tuason today at 12:00pm at the Google booth (#B206) to discuss maximizing quality while minimizing compute. #ICML2026

Attending #ICML2026? Stop by the Google booth (#B206) to attend a demo, chat with researchers, and learn about our groundbreaking work across the field of machine learning. Find information about our conference engagement as well as the schedule of booth activities at t.co/rkrvNZl2Nr

Google is proud to be a Diamond Sponsor of the 2026 International Conference on Machine Learning (#ICML2026), kicking off on July 7th in Seoul, Korea! Attending @icmlconf? Come by our booth (#B206) to learn more about our work, including papers, workshops, and booth activities at t.co/rkrvNZl2Nr

Introducing TabFM, a foundation model designed specifically for tabular data classification & regression. This approach allows generation of high-quality predictions on previously unseen tables in a single forward pass. Learn more and try out the model →goo.gle/4eR7uku

Today we announce the expansion of our Heat Resilience data to cover 50+ global cities. By fusing satellite data, high-resolution imagery, and AI, we map building-level rooftop reflectivity to help urban planners mitigate extreme heat. Learn more: goo.gle/4aq5rlZ

Today on the blog we introduce a method to retrofit Multi-Token Prediction onto frozen production models, accelerating on-device inference without the inefficiencies of separate drafters. Learn more →goo.gle/4uS4hH7

Read all about how linear elastic caching minimizes total cache cost by framing page eviction as a ski rental problem, using lightweight machine learning to optimize the trade-off between memory footprint and cache misses. Check out the blog →goo.gle/4oONMdt

Google researcher Dmitrii Kochkov was recently featured in @sciam's inaugural Young American Scientists list, as one of 28 early-career researchers whose work is shaping the future of science and society. He helped build the seasonal forecasting models that allowed Indian farmers to predict monsoons up to one month in advance — which could help ensure food security as climates change. #SciAmYoungScientists t.co/QFBErFdsNy

Check out the article: scientificamerican.com/article/dmitri…

At yesterday’s AI for the Planet event, we shared the progress we’ve made toward building AI tools to support our vision for crisis resilience efforts: a world where no one is surprised by a natural disaster. We’ve made critical information accessible via Google products that are used by billions, and partnered with governments and organizations around the world to help communities prepare for and respond to these crises.

Read more about our progress in the blog by @ymatias, GM of Google Research & Google Crisis Resilience Lead: goo.gle/3SJ928x

Today we present a study on how reasoning unlocks parametric knowledge in LLMs. We identify two key driving mechanisms, a computational buffer effect and factual priming, and suggest ways that can help build more reliable models. Learn more: goo.gle/44taazO

At yesterday’s AI for the Planet event we shared our north star for crisis resilience: no one should be surprised by a natural disaster. We’re leveraging AI, building helpful product features and collaborating with global partners to keep communities around the world safe and resilient in the face of natural disasters.

Read more about our progress in the blog by @ymatias, GM of Google Research & Google Crisis Resilience Lead: goo.gle/3SJ928x

New research published in @Nature with @GoogleDeepMind shows how AMIE utilizes long-context capabilities of Gemini models to ground its clinical reasoning in authoritative medical knowledge. In a multi-visit study with patient actors, AMIE demonstrated physician-level capabilities in longitudinal management reasoning, while scoring significantly higher in plan preciseness and guideline alignment. See the full story from @GoogleForHealth below 🧵↓

Today we introduce a new vectorized dataset for mapping fine-scale ecological features, such as hedgerows, that often go undetected by standard satellites. This precision provides a new roadmap for addressing climate & biodiversity challenges without compromising food security. More: t.co/NRU3hrqvrd

"I have yet to find a problem that is too difficult to solve. It's a question of how much effort you need to put into that." — @ymatias On the latest Release Notes, Yossi, VP, Google & GM, Google Research, sat down with @OfficialLoganK to discuss how Google teams drive

Vídeo

You can watch the full episode on YouTube: youtube.com/watch?v=FPBwad…

Today on the blog, we discuss a pathway for the second life of phones through the exploration of “phone cluster computing”, which can directly reduce the environmental footprint of computing by avoiding the need for further raw material extraction. More →goo.gle/4aJe5vO

GIF

The golden age of research ✨ x.com/ymatias/status…

Presenting recently published findings on how dermatology AI tools may help laypeople with their own skin-related questions, such as increased ability to name conditions, & some benefits in determining what next steps to take. Read more →goo.gle/3S5rNmv

🚀 Introducing Gemini-SQL2, our breakthrough text-to-SQL capability powered by Gemini 3.1 Pro! We've achieved state-of-the-art results on the highly competitive BIRD benchmark, translating natural language into execution-ready SQL queries. 🧵👇

📊 Data subtlety & complex business contexts make generating accurate SQL from natural language notoriously hard. Per the BIRD benchmark, which measures execution-verified accuracy, GeminiSQL-2’s SQL doesn't just look right, it also runs successfully.

Presenting Regularized f-Divergence Kernel Tests, a new framework for auditing machine unlearning and differential privacy. This method is more sensitive to localized data shifts and requires fewer samples than traditional tools. Learn more: goo.gle/4uoeZ8h

Congratulations to the authors of "Efficiently Reconstructing Dynamic Scenes One D4RT at a Time", recipient of the #CVPR2026 Best Paper Award! 🏆 arxiv.org/pdf/2512.08924 @GoogleDeepMind

It's the final day of #CVPR2026! Don't miss your chance to check out the Discover Android XR kiosk from 12:00 - 1:00 PM. 🕶️ x.com/GoogleResearch…

See Project Astra 3D at the #CVPR2026 Google booth #557 at 5:30pm! The team presents 3DCodeBench, demonstrating Gemini models' proficiency in generating diverse 3D objects through code execution. Presented by Lei Shu & Yipeng Gao. 3dcodebench.com @GoogleDeepMind

Missed the D4RT demo? Catch us at the kiosk today from 4:30 - 5:30 PM to learn about 4D scene reconstruction! 🎯 #CVPR2026 @GoogleDeepMind x.com/GoogleResearch…