Jieneng Chen

Huge congratulations to @jieneng_chen and collaborators on being selected for an oral presentation (top 6% of submissions!) -- very well deserved! This is exciting work, and I hope it gets the attention it truly deserves.

@vishalm_patel @jhuclsp @HopkinsEngineer Congrats!!

Honored to be named an IEEE Fellow for contributions to image processing, computer vision & biometrics. Also grateful to be an AAAI Senior Member and a 2025 Clarivate Highly Cited Researcher. Huge thanks to my students, mentors & collaborators! @jhuclsp @HopkinsEngineer

@shengyangzhuang great question. rn we’re focusing on monocular video — it’s more challenging but also much more common in the real world. That said, extending to multi-view setups is definitely feasible.

@jieneng_chen Looks cool! Does it work for multi view videos for an animal to reconstruct its 4D mesh?

A meaningful step forward for pet science. 🐾 Pets are our companions, yet modeling them in 3D is extremely hard due to limited data compared to humans. Thrilled to share that our 4D-Animal project is now online! We reconstruct animatable 3D animals from videos without requiring sparse keypoint annotations. paper: arxiv.org/pdf/2507.10437 code: github.com/zhongshsh/4D-A… Led by Shanshan Zhong (now a PhD student at CMU LTI) — a great journey exploring 4D vision together.

@mangahomanga Thanks, Homanga! Really appreciate it.

@jieneng_chen Cool stuff, congrats!!

Also, when it comes to world models for embodied agents, visual realism doesn’t necessarily translate to functional intelligence. You can explore our recent closed-loop evaluation here: world-in-world.github.io

The field has long been obsessed with judging World Models by their visuals—but that misses the point. In @jieneng_chen’s 𝐖𝐨𝐫𝐥𝐝-𝐢𝐧-𝐖𝐨𝐫𝐥𝐝, we propose the first closed-loop benchmark that compares WMs by 𝐞𝐦𝐛𝐨𝐝𝐢𝐞𝐝 𝐬𝐮𝐜𝐜𝐞𝐬𝐬, rather than 𝐯𝐢𝐬𝐮𝐚𝐥 𝐚𝐩𝐩𝐞𝐚𝐥. This is a major step toward 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛𝑎𝑙 evaluation of world models.

Are SOTA video models good world models for embodied agents? We present a benchmark evaluating this. We find that: - Visuals ≠ utility - Action data > bigger models - Test-time planning improves accuracy - Lots of room for improvement for all video models

🤯 Think better visuals mean better world models? Think again. 💥 Surprise: Agents don’t need eye candy— they need wins. Meet World-in-World, the first open benchmark that ranks world models by closed-loop task success, not pixels. We uncover 3 shocks: 1️⃣ Visuals ≠ utility 2️⃣ Action data > bigger models 3️⃣ Scaling test-time compute = more success 🤗 huggingface.co/papers/2510.18… 🌍 world-in-world.github.io 📄 arxiv.org/abs/2510.18135 github.com/World-In-World…

Dr. Kate Saenko @kate_saenko_ from @Meta will give a keynote talk titled with “Blind Spots in Multimodal AI Models.”

Dr. Fei Xia @xf1280 from @GoogleDeepMind will give a keynote titled with“Gemini Robotics 1.5: Generalist Robots with Advanced Embodied Reasoning, Thinking and Motion Transfer.”

🌺 Learn the next frontier of intelligence? Join us for the @ICCVConference workshop Embodied Spatial Reasoning (esr-2025.github.io) 🗓️ Oct 20 | 9:00–12:00 📍 Hawaii Convention Center, BallRoom A Featuring leading voices: @Sifei30488L @xiaolonw @xf1280 @kate_saenko_

@steph_milani @JHUCompSci @NYU_Courant Congrats! I really enjoyed your talk in Malone.

Another life update!! 🎉 I’m joining @JHUCompSci as an Assistant Professor starting Fall 2026! Apply to work with me on reinforcement learning, foundation models, & human-centered AI. Let’s build better AI agents 🤖🙆‍♀️🦀 Before that, I’ll join @NYU_Courant as an Assistant Professor/Faculty Fellow. Excited to spend a year in NYC!

@xiaolonw Congratulations!

Got my tenure! Very grateful to my students and collaborators.