If so, we want to hear from you! To apply, please follow these two steps: 1️⃣ Fill out the form below to personally express your interest to me: https://t.co/3OZPtCBDmI 2️⃣ Fill out the Google DeepMind "Student Researcher 2026" application form: https://t.co/nJYX3GrB2d

I am hiring a Student researcher at Google DeepMind for Winter/Spring or Summer of 2026! 🛠️ Interested in tool-use and reasoning for LLMs? 🚀 Up-to-date with frontier research? 🔬 Love to understand code & algorithms from first principles? 📚 Currently studying for a BS/MS/PhD?

After a year of team work, we're thrilled to introduce Depth Anything 3 (DA3)! 🚀 Aiming for human-like spatial perception, DA3 extends monocular depth estimation to any-view scenarios, including single images, multi-view images, and video. In pursuit of minimal modeling, DA3

Thanks to the team: @maxjiang93 , @stan188249301 , @tanmingxing, @AnguelovDrago, @hongjjeonS, Sakshum Kulshrestha, Yijing Bai, Jing Luo. - SceneDiffuser++ Paper:  https://t.co/HI7UaNGYo6 - Watch our video:  https://t.co/8wEX8bgsiW - SceneDiffuser Paper:  https://t.co/akpeaocCja

We propose a new task, CitySim, where given a city map and an AV software stack, the simulator can simulate the trip from point A -> B by populating the city around the AV and controlling all aspects of the scene (e.g., vehicles, pedestrians, traffic light states).

Though conceptually simple, this requires diffusion to learn to generate sparse tensors without prespecified sparse structure. During inference, we develop new clipping techniques to account for invalid entries in the denoising process.

For agents and traffic lights scene tensors, with a varying number of elements and feature dimensions, we can project scene tensors to the same latent dimension, and concatenate into a multi-tensor. We then pass this to a transformer denoiser backbone.

Like SceneDiffuser, we learn an agents "scene tensor," but generalize this to multi-tensor diffusion. Agent spawning, removal and occlusion can be jointly modeled simply via predicting an additional validity channel along with other agent features such as x, y, size, type, etc.

Only learned simulators can emulate the realism of crowded city scenes. Without the ability to insert or delete objects, these simulators can only simulate a few seconds before the scene becomes empty as initial logged agents leave the periphery of the AV.

Led by our amazing intern at Waymo Research, @stan188249301, SceneDiffuser++ is a diffusion model that is solely trained on the diffusion denoising objective, yet supports all insertion, deletion, and behavior control capabilities via simple autoregressive rollout.

Can a single autonomous driving simulation world model jointly insert, delete, and control the behavior of all agents and traffic lights in a bird's-eye-view scene? For the first time, we show this is possible in SceneDiffuser++, our CVPR '25 paper, w/ 60+ second simulations.🧵

Our Gemini 2.5 tech report ✨ is finally out on arXiv -- grateful for the collaboration with brilliant colleagues!

Great overview on the parallelism techniques needed for LLM training/inference from the new CS336 class at Stanford (Tatsunori Hashimoto & Percy Liang): https://t.co/9xxn9stCWq. Info-dense and doesn't shy away from the technical details!

More than 5x growth in paid trips from this time last year — each ride contributing to safer streets and more accessible mobility for all. Here’s to millions more!

Thrilled to share that the amazing work of @stan188249301 (@maxjiang93's and I's intern at Waymo Research last summer) is accepted to CVPR 2025! SceneDiffuser++ fulfills a vision we've had in mind for several years, and we weren't sure if it was even possible. More details soon!

🥁Introducing Gemini 2.5, our most intelligent model with impressive capabilities in advanced reasoning and coding. Now integrating thinking capabilities, 2.5 Pro Experimental is our most performant Gemini model yet. It’s #1 on @lmarena_ai leaderboard. 🥇

2️⃣ Fill out the Google DeepMind "Student Researcher 2025" application form: https://t.co/XauuxzBZ84 *this role is onsite in Mountain View ☀️🌊

🚀 To apply, just follow these two steps [please **do not** contact me on X, I don't check the inbox often.] 1️⃣ Fill out the form below to personally express your interest to me and my team: https://t.co/adyK5E0G8f

I am hiring a Student Researcher at Google DeepMind for 2025! 👩🔬 Interested in improving multi-turn optimization and reasoning capabilities of LLMs? 🧑‍🎓 Currently studying for a Bachelor's/Master's/PhD? 🧑‍💻 Have solid engineering and research skills? 🌟We want to hear from you!

I am sincerely grateful to have been able to learn, grow, publish, mentor exceptional interns, and together build the Waymo Driver these past few years. Leaving behind exceptional colleagues is bittersweet, but I’m excited for what lies ahead. ✨✨