🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Checkout Tianwei”s fast autoregressive video diffusion. A promising step towards real time interactive video generation!

An image of an object tells more than the visual geometry of objects —it’s also a physical snapshot of an object in a state of static equilibrium. Can we use that cue to get more information about the objects? Checkout Minghao’s work on this topic: PhysComp!

The slide is bad, her response to an audience is even worse… “Maybe there is one, maybe they are common, who knows what. I hope it was an outlier."

Unbelievable... This is explicit racial bias. How could this happen in NeurIPS? How could this be spoken by a top university professor, an invited keynote speaker?

🎉Excited to introduce our latest work, LinGen: Towards High-Resolution Minute-Length Text-to-Video Generation with Linear Computational Complexity! ✨For the first time, we demonstrate high-resolution 68-second video generation at 16fps on a single GPU— without relying on

💥 Think more real data is needed for scene reconstruction? Think again! Meet MegaSynth: scaling up feed-forward 3D scene reconstruction with synthesized scenes. In 3 days, it generates 700K scenes for training—70x larger than real data! ✨ The secret? Reconstruction is mostly

(Linear) Attention Mechanisms as Test-Time Regression By now, you've probably already heard of linear attention, in-context learning, test-time scaling, etc... Here, I'll discuss: 1. The unifying framework that ties them all together; 2. How to derive different linear

Introducing “Diffusion Autoencoders are Scalable Image Tokenizers” (DiTo). We show that with proper designs and scaling up, diffusion autoencoders (a single L2 loss) can outperform the GAN-LPIPS tokenizers (hybrid losses) used in current SOTA generative models. (1/4)

Very interesting work from MIT office mates! Diffusion Forcing with History Guidance introduces a novel approach to video generation, excelling at ultra-long sequences—800+ frames shown in the paper!

Tired of slow diffusion models? Our new paper introduces f-distill, enabling arbitrary f-divergence for one-step diffusion distillation. JS divergence gives SOTA results on text-to-image! Choose the divergence that suits your needs. Joint work with Weili Nie Arash Vahdat 1/N

Amazing results!

Halfmoon is Reve Image — and it’s the best image model in the world 🥇 (🔊)

Super excited to share that I’ve officially defended my PhD, wrapped up an incredible journey at Massachusetts Institute of Technology (MIT) and Adobe Research, and joined Reve! Thrilled to be working alongside the same amazing founders I teamed up with back in the Adobe days. That experience gave me deep

🔥Want to capture 3D dancing fluids♨️🌫️🌪️💦? No specialized equipment, just one video! Introducing FluidNexus: Now you only need one camera to reconstruct 3D fluid dynamics and predict future evolution! 🧵1/4 Web: yuegao.me/FluidNexus/ Arxiv: arxiv.org/pdf/2503.04720

An LLM generates an article verbatim—did it “train on” the article? It’s complicated: under n-gram definitions of train-set inclusion, LLMs can complete “unseen” texts—both after data deletion and adding “gibberish” data. Our results impact unlearning, MIAs & data transparency🧵

🔥Spatial intelligence requires world generation, and now we have the first comprehensive evaluation benchmark📏 for it! Introducing WorldScore: Unifying evaluation for 3D, 4D, and video models on world generation! 🧵1/7 Web: haoyi-duan.github.io/WorldScore/ arxiv: arxiv.org/abs/2504.00983

Excited to attend #ICLR2025 in person this year! I’ll be presenting two papers: 1. LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias 🔹 Oral Presentation: Session 3C (Garnet 216-218) — Apr 25 (Fri), 11:06–11:18 a.m. 🔹 Poster: Hall 3 + Hall 2B, Poster #593 — Apr