Google Research @GoogleResearch

How do goods travel 450 miles overnight? It’s thanks to careful logistics optimization, particularly in the middle-mile segment, but optimizing these networks is difficult without public data. MilleMiglia provides a standardized benchmark, using spatial clustering and gravity models to simulate realistic scenarios for middle-mile delivery problems. Learn more: goo.gle/4cTj1zy

"The best way to predict the future is to invent it." In a deep dive with VP, Google & GM, Google Research @ymatias on The Google Research Podcast, @sineadbovell explored how AI is transforming science & interfaces: 🔹 Ambient AI & GenUI: Interfaces that adapt in real time to how you think and learn. 🔹 Virtual Labs: Tools like AI Co-Scientist turning multi-year hypothesis generation into 3-day sprints. 🔹 Scientist as Architect: AI gives every student & researcher the resources they need to focus on asking better questions and exercising critical judgment. AI isn't replacing the researcher — it's an amplifier for human curiosity and judgment. 🔬⚡ 👉 Tune in to our pilot episode to hear the full conversation: youtube.com/watch?v=ggPWZ3…

Today we explore the future of practice. We harness generative UI with learning design guardrails to give teachers the ability to generate guided, learning interactive simulations — for every topic & student. Read more + check out the sample library with 30+ STEM interactives →goo.gle/46A7DF8

Spanning over 20 years, check out the latest in our AI language research in the 🧵⬇️

Yossi Matias @ymatias

4 days ago

Google’s work in language research spans over 20 years, beginning in 2006 with Google Translate. Today, these technologies power interactions in over 300 languages, reaching more than 7 billion people—86% of the global population. However, translation and text processing alone

2 1 24 15K 9

Introducing Retrieve-for-Train: a framework that accelerates complex AI search by replacing heavy autoregressive inference with a lightweight diffusion model. This allows for instant, expert-level search slates at a fraction of the cost. Learn more: goo.gle/4xuV3lB

Today, we are proud to share the new white paper, “How Google AI is reshaping crisis resilience,” bylined by @ymatias and Kate Brandt. It outlines how we’re building state-of-the-art AI models and working with global partners to predict natural disasters earlier and provide warnings so communities can take action before crises strike. The vision? A world where no one is surprised by a natural disaster. A few highlights: Riverine Floods: Forecasting up to 7 days advance across ~150 countries. Partners like GiveDirectly translate early warnings into anticipatory aid, such as direct cash transfers before floodwaters hit. Urban Flash Floods: Predictions up to 24 hours in advance using our Groundsource methodology powered by Gemini. Cyclones: Providing forecasters with an extra day of critical lead time. Wildfires: Tracking active fire boundaries across 34 countries and making the information available on Search and Maps. Read the white paper: goo.gle/3Uuzu6U

Can AI build new knowledge from experience? Join Google researchers Dima Damen and Viorica Patraucean alongside Yilun Du for the #ECCV2026 Panel today at 10:00am CEST in the Arena Room as they discuss the future of AI learning loops: eccv.ecva.net/virtual/2026/p… @GoogleDeepMind

If you missed the earlier session, join Gabriele Berton and Ye Xia today at 12:00pm CEST at the Google booth (#2) for a Demo & Q&A on TIPSv2 and spatial vision-language pretraining. Explore more: gdm-tipsv2.github.io #ECCV2026

Google Research @GoogleResearch

a week ago

Discover how enhanced patch-text alignment advances vision-language pretraining at #ECCV2026. Join Gabriele Berton and Ye Xia at the Google booth (#2) today at 1:00pm CEST for an interactive showcase of TIPSv2. Check out the paper and demo: gdm-tipsv2.github.io @GoogleDeepMind

5 12 117 22K 34

Introducing ToolGrad: an efficient framework for generating tool-use datasets. By generating ground-truth tool-use chains before prompts, it achieves a nearly 100% pass rate and improved LLM tool-use performance. More: goo.gle/4xNKTNH

Congratulations to Google DeepMind Researcher, Christine Kaeser-Chen and the Fine-Grained Visual Categorization (FGVC) Challenge Series team on the PAMI Mark Everingham Prize at ECCV 2026! 🏆 Thank you for your lasting contributions to the computer vision community. Well deserved! 👏

Interested in research opportunities at Google? Join our Google recruiting team today at 4:30 PM CEST at the Google booth (#2) for a Careers Spotlight Q&A! Learn all about Google Internships and the Student Researcher Program, and connect directly with our team. #ECCV2026

Discover how enhanced patch-text alignment advances vision-language pretraining at #ECCV2026. Join Gabriele Berton and Ye Xia at the Google booth (#2) today at 1:00pm CEST for an interactive showcase of TIPSv2. Check out the paper and demo: gdm-tipsv2.github.io @GoogleDeepMind

Attending #ECCV2026? Join Mohit Goyal at the Google booth (#2) today at 12:00pm CEST for a demo and Q&A on 'Perception, Not Reasoning, Limits Video Spatial Understanding' to explore how spatial perception enhances to visual language models. Stop by to ask your questions! split-spatial.github.io

Welcome to #ECCV2026! 🌟 Google is proud to be a Diamond Sponsor at @eccvconf in Malmö, Sweden (Sept 10–12). Visit us at booth #2 to explore our latest research, connect with our team, and see live demos! Learn more: goo.gle/ECCV2026 #GoogleResearch

🚨 NEW LABS EXPERIMENT 🚨 Great minds t̵h̵i̵n̵k̵ play alike! Play with Putty is a collaborative vibe coding tool that lets you build tools and websites together in real time. 🫟 Ready for your imagination to go multiplayer? Join the waitlist at labs.google/playwithputty and share your feedback with us! US only, ages 18+

Introducing WeatherNext 3! Built by @GoogleDeepMind and @GoogleResearch, this is our most advanced & accurate AI weather model, and now includes real-time satellite data, hourly refreshes, higher resolution, precise precipitation forecasting, and clean energy variables. Read the blog: blog.google/innovation-and… Build with WeatherNext 3: developers.google.com/weathernext More about Google Earth AI for planetary intelligence: ai.google/earth-ai/

Today on the blog, we evaluate ways to improve cross-population genetic risk prediction. While transfer learning from European cohorts improves prediction in small populations, it degrades accuracy as target cohort sample sizes grow. Learn more: goo.gle/4zR9XFf

Introducing a complete map of the male fruit fly brain and central nervous system, created in partnership with @HHMIJanelia. With over 166,000 neurons, this is the largest cellular brain map to date. Learn more: goo.gle/4qQTtIW