Detecting symmetries, particularly in noisy data, is challenging. Our method uses Langevin dynamics to make spotting symmetry easier, even in complex and noisy real-world data!

There has been significant recent interest in methods that use random walks to solve PDEs. In a project to be presented at SIGGRAPH Asia (w/Ekrem Yılmazer and Delio Vicini), we investigated how to solve *inverse PDE* problems by differentiating such solvers.

Curious whether video generation models (like #SORA) qualify as world models? We conduct a systematic study to answer this question by investigating whether a video gen model is able to learn physical laws. Three are three key messages to take home: 1⃣The model generalises

We've released our paper "Generating 3D-Consistent Videos from Unposed Internet Photos"! Video models like Luma generate pretty videos, but sometimes struggle with 3D consistency. We can do better by scaling them with 3D-aware objectives. 1/N page: genechou.com/kfcw

I’m recruiting PhD students with computer vision, robotics, or ML experience! We especially encourage applicants from physics and related fields who want to explore AI for Science. Join us by applying to Computer Science and Engineering at Michigan's PhD program!

Dynamic 3D scenes in your browser 🤯 Powered by @ArthurBrussee's amazing Brush rendering engine (github.com/ArthurBrussee/…)

Sharing something exciting we've been working on as a Thanksgiving gift: Diffusion Self-Distillation (DSD), which redefines zero-shot customized image generation using FLUX. DSD is like DreamBooth, but zero-shot/training-free. It works across any input subject and desired

A common question nowadays: Which is better, diffusion or flow matching? 🤔 Our answer: They’re two sides of the same coin. We wrote a blog post to show how diffusion models and Gaussian flow matching are equivalent. That’s great: It means you can use them interchangeably.

[Hiring!] I am hiring multiple PhDs USC Thomas Lord Department of Computer Science USC Viterbi School for this cycle. If you're interested in scene representations, neural simulation, generative AI, and robotics, feel free to mention my name in your application (no need to email). For USC masters/undergrads who're

🚀 Introducing CausVid: Instant video generation that plays the moment you hit "Generate", while maintaining state-of-the-art quality! Project Page: causvid.github.io. More details in the long thread.

Do large multimodal models understand how to make dresses for your winter holiday party💃? We introduce AIpparel, a vision-language-garment model capable of generating and editing simulation-ready sewing patterns from text and images. Project page at georgenakayama.github.io/AIpparel/.

How can we use wide-FOV cameras for reconstruction? We propose self-calibration Gaussian Splatting that jointly optimizes camera parameters, lens distortion, and 3D Gaussian representations to directly reconstruct from a set of wide-angle captures. page: denghilbert.github.io/self-cali/

Introducing AIpparel #CVPR2026 2025 - the first multimodal foundation model for digital garments. georgenakayama.github.io/AIpparel/ 1/6

Most video models struggle to feel like real worlds. They forget what’s just out of view, slow down as videos get longer, or breaks causality. We think State Space Models are a natural fit for models with: 🧠 long-term memory across hundreds of frames ⚡ constant-speed

Really impressive work on real-time video generation! I’m a fan of the principle of closing the train-test gap!

For those at CVPR, Aditya will be presenting this poster tomorrow at 10:30 (Exhibit hall D, Poster #34). Come hear about why neural field derivatives are noisy, and how we resurrect image processing ideas for neural fields!

We've released all code and models for FlashDepth! It produces depth maps from a 2k, streaming video in real-time. This was a really fun course project inspired by discussions with Mohamed Abdelfattah and Guandao Yang and we look forward to presenting it at #ICCV2025. GitHub:

Combining the benefits of RL and SFT with on-policy distillation, a promising approach for training small models for domain performance and continual learning.