💡 New work: You might not need math data to teach models math reasoning. Recent 🔥 RLVR works challenge the need of *labels* of math questions. We find just playing video games, eg. Snake, can boost multimodal reasoning. No math *questions* needed. arxiv.org/abs/2506.08011🧵👇

To wrap up the open-sourcing season, we’re excited to announce that DiffusionRenderer, based on the NVIDIA Cosmos world model, is now open sourced! That means that the code of all three of our #CVPR25 oral papers is now available: - 3DGUT - DiffusionRenderer - Diffix3D+ #CVPR2026

🚀 Introducing Cosmos-Predict2! Our most powerful open video foundation model for Physical AI. Cosmos-Predict2 significantly improves upon Predict1 in visual quality, prompt alignment, and motion dynamics—outperforming popular open-source video foundation models. It’s openly

If you're at #CVPR2025 don't miss Yiqing Liang's talk on our scene flow estimation work, which was also a CVPR Best Paper Award nominee! Collaboration between NVIDIA AI and Brown University

Anthropic just dropped the beautiful explaination of how they built a multi-agent research system using multiple Claude AI agents. A MUST read for anyone building multi-agent system. A lead agent plans research steps, spawns specialized subagents to search in parallel, and

Honored to be interviewed by RSIP Vision and see our #CVPR2025 2025 Best Paper Award Candidate ZeroMSF featured on the front page! Thank you for spotlighting our zero-shot scene flow research! 🚀 More details from our website: research.nvidia.com/labs/lpr/zero_… #NVIDIA #CVPR2025

V-JEPA 2 Self-Supervised Video Models Enable Understanding, Prediction and Planning

🚀 Introducing UniRelight, a general-purpose relighting framework powered by video diffusion models. 🌟UniRelight jointly models the distribution of scene intrinsics and illumination, enabling high-quality relighting and intrinsic decomposition from a single image or video.

Preprint of (not) today: Harley et al., "AllTracker: Efficient Dense Point Tracking at High Resolution" -- alltracker.github.io Efficient architecture to track all points/pixels in real-time. Matches multiple frames in real time at high resolution via RNN-based refinement.

Tricky sample for AllTracker. It has never seen water at training time, so I'm glad it knows to quickly discard the tracks there... The bird that travels right-to-left (behind the wings most of the time) could be handled better.

🔥 𝗧𝗵𝗿𝗶𝗹𝗹𝗲𝗱 𝘁𝗼 𝘀𝗵𝗮𝗿𝗲 𝗼𝘂𝗿 𝗹𝗮𝘁𝗲𝘀𝘁 𝘁𝗿𝗶𝗹𝗼𝗴𝘆 𝗼𝗳 𝗽𝗮𝗽𝗲𝗿𝘀 𝗼𝗻 𝗔𝗜 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝗠𝗼𝗱𝗲𝗹𝘀, 𝗽𝘂𝘀𝗵𝗶𝗻𝗴 𝘁𝗵𝗲 𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝘁𝗵𝗲𝗼𝗿𝘆, 𝗱𝗮𝘁𝗮, 𝗮𝗻𝗱 𝗺𝗼𝗱𝗮𝗹𝗶𝘁𝘆: 🔴 𝚃̲𝚑̲𝚎̲𝚘̲𝚛̲𝚢̲ – 𝘋𝘪𝘴𝘊𝘖:

Brown Visual Computing is in the top 0.1% at #CVPR2026 2025, including research from BVC student Yiqing Liang and her collaborators at NVIDIA that was recognized as a best paper award candidate with oral presentation. Learn more: cs.brown.edu/news/2025/06/3…

Based on developments on "flow-map / average velocity" type methods, such as consistency trajectory models, shortcut models, IMM, and mean flow, I believe that the community will develop a proper replacement to diffusion / flow matching in 6 - 12 months.

📢Thrilled to share that I'll be joining Harvard and the Kempner Institute as an Assistant Professor starting Fall 2026! I'll be recruiting students this year for the Fall 2026 admissions cycle. Hope you apply!

This is Ray3. The world’s first reasoning video model, and the first to generate studio-grade HDR. Now with an all-new Draft Mode for rapid iteration in creative workflows, and state of the art physics and consistency. Available now for free in Dream Machine.

🎉 [CVPR 2025] ZeroMSF Code Release! 3D scene flow from a single camera, with no fine-tuning on new domains? That’s the challenge we tackled in Zero-MSF (Zero-shot Monocular Scene Flow). 💡 Motivation Scene flow captures both geometry and motion, but existing methods crumble

We build world models grounded in physical reality, naturally generalizing across cars and robots 🤖