Kaiyang Zhou
230 posts
@kaiyangzhou
Assistant Professor at HKBU. Interested in machine learning & computer vision.
🎙️The AI Talks 3D World Model for Robotics - by Wenlong Huang @wenlong_huang Can we teach robots to imagine the future before they act? From building a 3D intuition for physics with PointWorld to zero-shot synthesis of real-world manipulation skills, let's dive in. Join us👇
Today Ropedia releases Xperience-10M at #GTC day 1 — World largest real human 4D interaction dataset at 10M scale. Each trajectory aligns: • visual observations • spatial structure • human motion • interaction dynamics • task semantics A new foundation for physical and spatial AI, try it out @huggingface huggingface.co/datasets/roped…
One static model does not fit all😭 We just dropped our latest work: Functional Neural Memory. Instead of static models, we generate custom "parameters" for every single input. ✅Prompt your model anytime ✅Instant personalization ✅Better instruction following ✅Flexible & dynamic memory (w/o memory bank✌️) (🧵1/6)
Build a real-time AI assistant! 🎥 Streamo (to appear in #CVPR2026) enables turning any offline LLM into a live streaming video model. Key features: 💫 Real-time interaction ✨ Diverse instructions 🌟 Multi-task capable Code:github.com/maifoundations… Paper:arxiv.org/pdf/2512.21334
🎙️The AI Talks | S5E6 Introducing SAM 3D: Powerful 3D Reconstruction for Physical World Images — by Weiyao Wang @weiyaow1 & Xitong Yang. Can we "3D-fy" everything? 🌍🏃♂️ From universal objects to human bodies, explore the new SAM 3D family. Join us👇
Just read an excellent survey paper on AI agents. Measuring Agents in Production 📄 Paper: arxiv.org/pdf/2512.04123 🎬 Explained Video (Chinese): bilibili.com/video/BV1rkBNB… It gave me an epiphany about the real status of agents in production and prompted me to reflect on what makes a good survey paper. Here's my checklist for writing a good survey paper in 2026: ⭐ Novel insights – not just a literature dump ⭐ Unified understanding – synthesize, don't just summarize ⭐ Predictive vision – point to future directions that aren't yet consensus ⭐ Final acid test – if your paper could be written by frontier models with deep research, drop it.
📢 New launches @synvoAI 🍫Local Cocoa Offline, on-device💻 multimodal contextual memory for privacy-first🔒 use. 🔥 Open-source: github.com/synvo-ai/Local… 🌐Synvo API Plug-and-play multimodal contextual memory for AI workflows or agents 🤖 Try it: console.synvo.ai
