🚀UniWorld: a unified model that skips VAEs and uses semantic features from SigLIP! Using just 1% of BAGEL’s data, it outperforms on image editing and excels in understanding & generation. 🌟Now data, model, training & evaluation script are open-source! github.com/PKU-YuanGroup/…

MoVieS Motion-Aware 4D Dynamic View Synthesis in One Second

Amazing！

thanks for sharing!

ChatGPT can now do work for you using its own computer. Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.

早晨起来，意外发现 Qwen3 Coder 发布了。 Qwen3 Coder 一个具备 Agent 能力的代码模型。 这个模型在 Agentic Coding、Agentic Browser-Use 和 Agentic Tool-Use 上取得了开源模型的 SOTA。 简单说，代码和 Agent 能力，可以和 Claude Sonnet4 相媲美。 模型总参数量只有 480B，激活参数 35B。

🧠Video Thinking Test for Reasoning LLMs🧠 *Video Thinking Test* (📽️Video-TT📽️) is a holistic benchmark to assess the advanced reasoning and understanding correctness/robustness between LLMs and humans #ICCV2025 - Project: zhangyuanhan-ai.github.io/video-tt/ - Data: huggingface.co/datasets/lmms-…

Dataset Distillation as Data Compression: A Rate-Utility Perspective arxiv.org/abs/2507.17221 Read this paper tonight, get me some sense: Dataset Distillation ≈ Visual Tokenization? Dataset Distillation: Replace full dataset with few synthetic samples Visual Tokenizer: Replace

I tested ChatGPT-5 and Gemini 2.5 Pro with same critical prompts. The results will shock you. ChatGPT-5 Vs. Gemini 2.5 Pro (Video demos are included)

This guy literally dropped the best visual guide to LLMs you’ll ever see

Introducing LucidFlux-14B — caption-free image restoration for the real world. SOTA on 6 metrics, rivaling closed-source. Built on a 12B Flux-DiT with a unified dual-branch design + adaptive temporal/depth fusion; SigLIP preserves semantics without text.

🚀 SANA-Video: Linear Attention + Constant-Memory KV Cache = Fast Long Videos 💥 Key Features 🌟 🧠 Linear DiT everywhere → O(N) complexity on video-scale tokens 🧰 Constant-memory Block KV cache → store cumulative states only (no growing KV) 🔄 🎯 Temporal Mix-FFN + 3D RoPE

🚀 Introducing FlashI2V: The game-changer in Image-to-Video generation! 🔥 Solving conditional image leakage with Latent Shifting & Fourier Guidance. 1.3B parameters, outperforms CogVideoX1.5-5B in speed, quality & generalization. github.com/PKU-YuanGroup/…

To push self-driving into situations wilder than reality, we built a neural network world simulator that can create entirely synthetic worlds for the Tesla to drive in. Video below is fully generated & not a real video

Diff4Splat: Controllable 4D Scene Generation with Latent Dynamic Reconstruction Models Contributions: • We propose DIFF4SPLAT, a unified diffusion-based model that directly generates deformable 3D Gaussians for controllable 4D scene synthesis. • We construct a large-scale 4D

🦋DynamicVerse: A Physically-Aware Multimodal Framework for 4D World Modeling Excited to share that our work in the #NeurIPS2025 ! - A large-scale 4D + instance + semantics + caption dataset with 100K in-the-wild scenes, supporting 4D world modeling by combining classic 3D

Excited to see JarvisArt hit 700🌟! To support the community, we've released the full pipeline: 1️⃣ Multi-machine protocol (Agent ↔️ Lightroom) 2️⃣ Data construction scripts & MMArt-Bench 3️⃣ All training/inference/eval code Check it out & contribute! 🚀 🔗 github.com/LYL1015/Jarvis…

Edit designs, images, and documents. Adobe Express, Photoshop, and Acrobat are now in ChatGPT. adobe.ly/3XNhgvO

🎨Introducing JarvisEvo: The First Self-Evolving Photo Editing Agent! From tool-user to Creator. Moving beyond "blind" CoT to true visual perception & reflection. Project Page: jarvisevo.vercel.app Arxiv: arxiv.org/pdf/2511.23002 GitHub: github.com/LYL1015/Jarvis…