Yanjia Huang

[Yanjia Huang @Saam_0812]

From imitation to Spatial Reasoning

Learn2Fold is based on a simple idea: treat cloth folding as a robotics task, moving beyond imitation learning toward better generalization across deformable objects. We hope this is a real step toward reasoning.

[Yanjia Huang @Saam_0812]

extended submission to Early July！

[Yanjia Huang @Saam_0812]

Missed CVPR/ICRA? Join us for the 1st Embodied4Arts Workshop @ RSS 2026!

Robotics is more than picking, placing, and chasing higher VLA success rates.

We are looking for creative robotics work that explores how robots can paint, draw, perform, craft, fabricate, compose, collaborate, and participate in artistic practice.

We welcome work on robot painting, embodied music, robotic craft, performance robotics, human-robot co-creation, creative manipulation, deformable materials, embodied generative systems, and anything that asks what robots can create, express, or make with us.

Two tracks:
• Workshop Papers & Demos
• Robot Painting Challenge

Submission deadline: June 24, 2026
Workshop date: July 13, 2026
Location: RSS 2026, Sydney, Australia

Website: https://t.co/nxOWyQui52
OpenReview: https://t.co/5EVCrzHKAF

If your robot does something creative, expressive, strange, beautiful, or materially interesting, we want to see it.
#rss2026 #robotics #embodied

[Yanjia Huang @Saam_0812]

Missed CVPR/ICRA? Join us for the 1st Embodied4Arts Workshop @ RSS 2026!

Robotics is more than picking, placing, and chasing higher VLA success rates.

We are looking for creative robotics work that explores how robots can paint, draw, perform, craft, fabricate, compose, collaborate, and participate in artistic practice.

We welcome work on robot painting, embodied music, robotic craft, performance robotics, human-robot co-creation, creative manipulation, deformable materials, embodied generative systems, and anything that asks what robots can create, express, or make with us.

Two tracks:
• Workshop Papers & Demos
• Robot Painting Challenge

Submission deadline: June 24, 2026
Workshop date: July 13, 2026
Location: RSS 2026, Sydney, Australia

Website: https://t.co/nxOWyQui52
OpenReview: https://t.co/5EVCrzHKAF

If your robot does something creative, expressive, strange, beautiful, or materially interesting, we want to see it.
#rss2026 #robotics #embodied

[Yanjia Huang @Saam_0812]

@druncle @_vztu 你在叫你吗呢 你真做过research？真做过robo？看看google scholar？能把ap翻译成助教是不是只会用google trans？你什么水平？是家里户口本单开一页所以这样说话？说别人凤凰男不会是自己当不上然后在破防然后在这里酸？

[Yanjia Huang @Saam_0812]

@NicholasJBryan @AdobeResearch Hi Bryan，do u consider applying AI4music on robotics？

[Yanjia Huang @Saam_0812]

@lrzneedresearch 实则恋爱谈不上 research也做不好😭

[Yanjia Huang @Saam_0812]

@Patarino @Skiminok NeurIPS

[Yanjia Huang @Saam_0812]

@GuanyaShi @LeCARLab Hi Prof. Guanya, your dm isn’t open yet lol. Looking forward to seeing you soon!

[Chenfanfu Jiang @chenfanfujiang]

Pixels are the result. Strokes are the story.
We turn one image into vectorized Bézier strokes and differentiable smudges that could have painted it. Watch the canvas come alive: https://t.co/1JKW5ZBEzi

[Yanjia Huang @Saam_0812]

I am presenting the most expressive robotic piano playing policy at NeurIPS2025！DM is open！Happy to chat with everyone！#NeurIPS2025

[Zhengzhong Tu @_vztu]

🎶🤖 Can a Robot Play Piano with the Expressiveness of a Human? Meet 𝗣𝗔𝗡𝗗𝗢𝗥𝗔! 🎹🌟

We’re thrilled to introduce 𝗣𝗔𝗡𝗗𝗢𝗥𝗔, our innovative diffusion policy learning framework that enables dexterous robotic piano performances—combining technical precision with genuine musical expressiveness! 🎼✨

🎯 𝗪𝗵𝘆 𝗶𝘀 𝗣𝗔𝗡𝗗𝗢𝗥𝗔 𝗴𝗿𝗼𝘂𝗻𝗱𝗯𝗿𝗲𝗮𝗸𝗶𝗻𝗴?
🚀 Employs diffusion-based generative models (conditional U-Net with FiLM) for smooth, expressive finger trajectories.

🤖 Introduces a novel Large Language Model (LLM)-driven reward system for dynamic, semantic feedback on musical expression.

🎵 Features a unique composite reward, uniting task accuracy, audio fidelity, and high-level musical style guidance from an LLM oracle.

📊 𝗢𝘂𝘁𝘀𝘁𝗮𝗻𝗱𝗶𝗻𝗴 𝗥𝗲𝘀𝘂𝗹𝘁𝘀:
🥇 Achieves state-of-the-art performance on the ROBOPIANIST benchmark, significantly surpassing existing baselines in precision and expressiveness.

📈 Extensive experiments and ablation studies confirm the critical role of diffusion-based denoising and semantic reward shaping for artistic robotic manipulation.

🔗 𝗪𝗲𝗯𝘀𝗶𝘁𝗲 & 𝗱𝗲𝗺𝗼 𝘃𝗶𝗱𝗲𝗼𝘀: https://t.co/isPQQUjBJC
📄 𝗔𝗿𝘅𝗶𝘃: https://t.co/ErgGczaYdZ

Let’s open the box of robotic creativity—welcome to the age of expressive, dexterous robotic artists! 🎹✨🤖

[Zhengzhong Tu @_vztu]

🎉 [𝗡𝗲𝘂𝗿𝗜𝗣𝗦 𝟮𝟬𝟮𝟱] 𝗣𝗮𝗽𝗲𝗿 𝗔𝗰𝗰𝗲𝗽𝘁𝗲𝗱 𝗮𝗻𝗱 𝗖𝗼𝗱𝗲 𝗥𝗲𝗹𝗲𝗮𝘀𝗲!

Not everyone is a photo pro — even with advanced AI tools. Restoring an image to professional grade often requires either 𝘥𝘦𝘦𝘱 𝘱𝘩𝘰𝘵𝘰𝘨𝘳𝘢𝘱𝘩𝘺 𝘦𝘹𝘱𝘦𝘳𝘵𝘪𝘴𝘦 or 𝘣𝘳𝘰𝘢𝘥 𝘬𝘯𝘰𝘸𝘭𝘦𝘥𝘨𝘦 of how to combine many different AI models.

🙋 𝗦𝗼 𝘄𝗲 𝗮𝘀𝗸𝗲𝗱 𝗼𝘂𝗿𝘀𝗲𝗹𝘃𝗲𝘀: Can we build a restoration agent that fully automates this process — no domain expertise required?

To answer this, we @TAMU partnered with @Topaz and scholars from @Stanford, @Caltech, @UTAustin, @Snap, @ucmerced, @CUBoulder to develop the 𝘄𝗼𝗿𝗹𝗱’𝘀 𝗳𝗶𝗿𝘀𝘁 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗽𝗵𝗼𝘁𝗼 𝗿𝗲𝘀𝘁𝗼𝗿𝗮𝘁𝗶𝗼𝗻 𝘀𝘆𝘀𝘁𝗲𝗺.

Our system brings together over 𝟱𝟬 𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝗶𝘇𝗲𝗱 𝗔𝗜 𝗺𝗼𝗱𝗲𝗹𝘀 — for denoising, deblurring, upscaling, face recovery, detail enhancement, and more.

🔬 It diagnoses the input image.
🦾 It plans and executes an action graph (just like a human editor).
☑️ After each step, it evaluates the output and adapts the plan if needed.

The results are phenomenal — as you’ll see in the video.

We believe democratizing professional photography will empower anyone to create images suitable for professional use cases. And because we see this as a responsibility, we are 𝗳𝘂𝗹𝗹𝘆 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗶𝗻𝗴 𝘁𝗵𝗲 𝘀𝘆𝘀𝘁𝗲𝗺 — to accelerate progress in agentic photo restoration and support the broader community.

👉 𝗖𝗵𝗲𝗰𝗸 𝗼𝘂𝘁 𝘁𝗵𝗲 𝗰𝗼𝗱𝗲: https://t.co/ZJU6QcBcgK
📄 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝘄𝗲𝗯𝗽𝗮𝗴𝗲: https://t.co/7nTNyh03p4

𝗜'𝗺 𝗹𝗼𝗼𝗸𝗶𝗻𝗴 𝗳𝗼𝗿𝘄𝗮𝗿𝗱 𝘁𝗼 𝘀𝗲𝗲𝗶𝗻𝗴 𝘄𝗵𝗮𝘁 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗯𝘂𝗶𝗹𝗱 𝘂𝘀𝗶𝗻𝗴 𝗼𝘂𝗿 𝗼𝗽𝗲𝗻–𝘀𝗼𝘂𝗿𝗰𝗲 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗮𝗻𝗱 𝘁𝗼 𝗿𝗲𝘃𝗶𝗲𝘄𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗱𝗲𝗺𝗼 𝗮𝗻𝗱 𝗳𝗲𝗲𝗱𝗯𝗮𝗰𝗸!

[Yanjia Huang @Saam_0812]

今天在喜欢的湖边坐了好久。
到的时候天已经黑了，仿佛不是我认识的那个湖。
我坐着，看它一点点隐下去。
好像等我起身的时候，脚下会自己选一块石头落下去。
它终究也会慢慢失去形状。
好像会带走点什么，也会留下点什么。

[Tairan He @TairanHe99]

Two weeks ago I passed my PhD thesis proposal 🎉 Huge thanks to my advisors @GuanyaShi & @ChangliuL, my committee, and everyone who has helped me along the way.

Last week I also gave a talk at UPenn GRASP on our 2-year journey in humanoid sim2real—reflections, lessons, and outlook. Great Q&A, now online: 🔗 https://t.co/tiTbTeXsEW

During the visit, I had inspiring 1:1s with 10 professors and even an hour with GRASP Lab founder Ruzena Bajcsy. I asked her: after nearly 60 years as a robotics professor, what defines a truly great researcher? She said:
1️⃣ Don’t publish many papers
2️⃣ Don’t pick trivial problems

Two principles I’ll strive to follow for the rest of my PhD journey.

[Science Robotics @SciRobotics]

Scientists have designed a #ReinforcementLearning-based framework that enables multiple robot arms to perform up to 40 tasks simultaneously without colliding in a crowded workspace. @GoogleDeepMind

Learn more in Science #Robotics: https://t.co/B7HDD3lsIL

[Yanjia Huang @Saam_0812]

@ChongZitaZhang @SciRobotics @JayHe748646 @leggedrobotics what you are a real man bro?! I saw all the papers you share and thought you are an agent lol. Congrats!!!

[Embodied AI Reading Notes @EmbodiedAIRead]

Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation

Project: https://t.co/DI5wq4tpHB
Paper: https://t.co/YDdoE7Pkb4

New paper from AgiBot on World Model for manipulation.
- GE-Base: a video diffusion model trained on 3000hours, over 1million manipulation episodes from AgiBot-World-Beta dataset.
- GE-Act, a flow-matching action model using the latent visual features from GE-Base.
- GE-Sim enables a world neural simulator for closed-loop control and evaluation.
- The authors claim to open source all code, models and benchmarks.

[Yanjia Huang @Saam_0812]

@_vztu @grok Ur absolutely more handsome than that lol

[David McAllister @davidrmcall]

Excited to share Flow Matching Policy Gradients: expressive RL policies trained from rewards using flow matching. It’s an easy, drop-in replacement for Gaussian PPO on control tasks.

[Yanjia Huang @Saam_0812]

Dolce far niente