VGGT has been re-licensed to allow commercial usage. Enjoy the gift 😉😇 👉 huggingface.co/facebook/VGGT-…

We just shipped Gemini 2.5 Deep Think it doesn't just recall research papers - it fuses ideas across papers in ways I haven't seen before this level of capability demands careful evaluation model card below 👇

All week during rebuttals, I have started each day with the same reminder: stay humble, stay kind, don't let this turn me mean. When I was doing PhD, reviewers never felt this mean. There is a bright-eyed student sitting on the other side, and such reviews will destroy

Here we go! 🚀🚀🚀

I am happy to share the work of our team. The outcome of a collaborative effort, by a joyful group of skilled and determined scientists and engineers! Congrats to the team on this amazing milestone!

🏆🏆🏆Clash of the Titans (GPT 5 vs. Gemini 2.5 Pro) on GAMEBoT: Connect4-->11:8 Checkers-->20:0 #GPT5,#Gemini

A complete game of Checkers (in GAMEBoT): GPT-5 vs. Gemini 2.5 Pro

🚨🚨🚨Deadline <10 days 🚨🚨🚨 Submit your extended abstracts on AI driven comic analysis to the COMIQ workshop #ICCV2025 comiq-iccv25.github.io

🚀 Excited to share that our paper "𝗥𝗲-𝗔𝗹𝗶𝗴𝗻: 𝗔𝗹𝗶𝗴𝗻𝗶𝗻𝗴 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝘃𝗶𝗮 𝗥𝗲𝘁𝗿𝗶𝗲𝘃𝗮𝗹-𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗗𝗶𝗿𝗲𝗰𝘁 𝗣𝗿𝗲𝗳𝗲𝗿𝗲𝗻𝗰𝗲 𝗢𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻" has been accepted to EMNLP 2025 (Main Track)! 🎉 Large

amazing!

Code released for Hyperbolic Category Discovery --- A simple hyperbolic framework for learning hierarchy-aware representations and classifiers for GCD, achieving the SOTA performance. 💻 Code: github.com/Visual-AI/HypCD 🌐 Page: visual-ai.github.io/hypcd/

visited Oxford for a couple of days for the robotics research group’s 40th anniversary (which VGG is part of) and its founder Mike Brady’s 80th birthday. it was nice to be back, walking in the University Parks and seeing dearly missed friends. some advice I heard there & loved:

Introducing Inpaint4Drag, a novel drag-based editing framework, which is fast and effective.🎨🚀🛠️ 📜Paper: arxiv.org/abs/2509.04582 🧑‍💻Project&code: visual-ai.github.io/inpaint4drag/ #ICCV2025

RIP🕯️🕯️🕯️

Our paper 'Self‑Labelling via Simultaneous Clustering and Representation Learning' just got its 1000th citation. On that occasion, I want to give my perspective on this question: Who or what is Sinkhorn–Knopp? Short answer: It’s the little ~1960s matrix‑normalization workhorse

three years ago, DiT replaced the legacy unet with a transformer-based denoising backbone. we knew the bulky VAEs would be the next to go -- we just waited until we could do it right. today, we introduce Representation Autoencoders (RAE). >> Retire VAEs. Use RAEs. 👇(1/n)