iiinova — where your digital self keeps evolving

Chen Geng @chen-geng.bsky.social · 01/06/2026

… or simply in virtual spaces with artist-created or AI-generated 3D assets! (9/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

… or turning @guangzhao-he.bsky.social’s Cornell office into a playground (8/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

… or an embarrassingly messy kitchen like ours :) No special setup. Just a phone scan. (7/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

Pass an object mesh into the model, solve the ODE, and you get its 4D dynamics. That’s the whole pipeline. So you can scan a real room and start interacting with the objects in it. Here’s me interacting with objects in a kitchen on the second floor of Gates Building at Stanford. (6/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

NeuROK learns an object’s configuration space as a latent space from large-scale 4D data. Then we simulate on the learned manifold with the Euler–Lagrange equations. One ODE for all object types. Interactive illustration: chen-geng.com/neurok/#method (5/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

The magic comes from Lagrangian mechanics: the right coordinates can make hard dynamics much simpler. Instead of tracking redundant coordinates and enforcing constraints afterward, NeuROK learns a compact state space where constraints are built in from data. (4/n)

110

Chen Geng @chen-geng.bsky.social · 01/06/2026

Traditional simulation often means picking models, preparing geometry, and tuning parameters. NeuROK is one general framework for articulated objects, cloth, elastic bodies, and multi-body systems — no per-category tuning, no separate models. (3/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

NeuROK takes a static 3D mesh — with no physical or structural annotations — and generates the full 4D trajectory of every vertex under given physical conditions. No material parameters. No category labels. Geometry in, motion out. (2/n)

100

Chen Geng @chen-geng.bsky.social · 01/06/2026

🌟Your 3D world models are now alive and interactable! 🚀Introducing NeuROK, a neural simulation framework that turns any static 3D object into an interactive 4D asset — no per-category physics, no physical annotations for training. 📄 chen-geng.com/neurok 🧵 1/n

121

Reposted by Chen Geng

Chen Geng @chen-geng.bsky.social · 16/01/2026

Introducing CHORD, a universal framework for generating scene-level 4D dynamic motion from any static 3D inputs. It generalizes surprisingly well across a wide range of objects and can even be used to learn robotics manipulation policy! Project page: yanzhelyu.github.io/chord. a 🧵: 1/n

143

Chen Geng @chen-geng.bsky.social · 16/01/2026

Yanzhe (yanzhelyu.github.io) is applying for PhD programs this cycle. He is truly exceptional, and any top program would be fortunate to have him. Don’t miss out! 🧵 15/15

yanzhelyu.github.io

Yanzhe Lyu

000

Chen Geng @chen-geng.bsky.social · 16/01/2026

This project is led by our talented summer intern Yanzhe Lyu (yanzhelyu.github.io), in collaboration with Karthik Dharmarajan, @zhang-yunzhi.bsky.social, Hadi Alzayer, @elliottwu.bsky.social, and @jiajunwu.bsky.social. 🧵 14/n

yanzhelyu.github.io

Yanzhe Lyu

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

This thread only scratches the surface. 🧊 We have a massive gallery of generated 4D results, comparisons, and interactive demos on our website. 🔎 Explore the full gallery: yanzhelyu.github.io/chord/#app 🧵 12/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

Ezio is playing with his eagle. 🦅🖐️ This 4D scene demonstrates our model's ability to generate human-animal interactions. Notice the natural timing and coordination between the two figures. 🧵 11/n

110

Chen Geng @chen-geng.bsky.social · 16/01/2026

Even the Dark Side has to obey the laws of physics! 🌑 Watch a 4D scene of Darth Vader pushing down on a desk lamp. Notice the articulation: the lamp folds and compresses naturally under the force of the hand. 💡📉 🧵 10/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

A cat jumping onto a cushion. 🐈🛋️ Notice the realistic physical interaction: Watch how the cushion deforms and reacts to the cat's weight as it lands. Our pipeline generates these subtle contact dynamics automatically, without complex physical equations defined in simulators! 🧵 9/n

110

Chen Geng @chen-geng.bsky.social · 16/01/2026

With CHORD, the creative possibilities are endless. 🎨 Users can generate a limitless variety of fun 4D worlds. 🍿 Watch this: A generated 4D scene of Captain America taking a break to pet a dog! 🛡️🐕 🧵 8/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

How does it work? We employ a robust distillation strategy to learn dynamic motion patterns from large-scale video generative models. Read more in our paper: yanzhelyu.github.io/chord/static... 🧵 7/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

Explicit 4D representations = Perfect memory for 4D world generation. 🧠 This allows us to generate coherent, minute-long 4D worlds without the physics breaking down. Witness a full minute-long sequence: 1. Move plate to microwave 🍽️ 2. Wait for heating ⏳ 3. Retrieve plate ✅ 🧵 6/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

By leveraging explicit 3D representations, CHORD generates motion that is physically grounded, not just visually plausible. This ensures physical consistency, enabling direct transfer to robotic systems. 🦾 🔗 View more examples here: yanzhelyu.github.io/chord/#robot 🧵 5/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

Our pipeline generates spatially consistent 4D scenes, meaning they can be viewed from any angle. 🔄 👇 Explore the interactive viewer yourself: yanzhelyu.github.io/chord/#viewer 🧵 4/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

Crucially, our pipeline operates without any category-specific priors or dynamic structural annotations. This flexibility allows CHORD to generate a diverse range of complex object interactions. 👇 Watch the example below: A generated 4D world showing a child launching a brick with a seesaw. 🧵: 3/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

GenAI excels at generating static 3D shapes, but 4D World Models demand realistic dynamics. CHORD bridges this gap: Input: Static 3D objects (zero dynamic annotations). Output: Plausible 4D object motion & coherent 4D scenes. No rigs or skeletons required. 🧵: 2/n

100

Chen Geng @chen-geng.bsky.social · 16/01/2026

143