Monoform now turns your photos and scans into sim-ready scenes for robotics.
Our 3D models and agents reconstruct scene geometry and object placement automatically, so you can get to testing without hours of manual cleanup.
These scenes took ~10 minutes to build.
Try it here: monoformlabs.dev/
Agentic real2sim2real is going to unlock many long-horizon capabilities for robotics. In this work led by Xudong, we aim to make agent-generated assets more "physically correct", with less self-penetration and better performance in downstream robotic tasks!
A couple of questions at the top of my mind:
1. Two obvious limitations of current agentic real2sim are unrealistic materials and lack of metric-accurate object scale. How much do these actually matter for downstream robotic evaluation and post-training?
2. If the goal is to improve a policy in a specific target environment, e.g., through post-training in an agent-crafted simulation, what's the right balance between diversity (generating many scene variations) and faithfulness (closely matching the target environment)?
3. Physical properties are hard to get exactly right with vision alone. What is the right interface to enable models to absorb interaction data (think end effector trajectory, force, contact, etc.)? Maybe coding (writing code to digest the raw data) would already be enough?
Looking forward to seeing more work on this front!
How can we generate and evaluate physically grounded articulated 3D assets?
🚀 Introducing ARTISAN: a multi-agent generation pipeline and a comprehensive benchmark
🔗 Project: physicalvision.github.io/art…
💻 Code and datasets: github.com/physicalvision/Ar…
🚀 PhysAI is coming to ECCV 2026!
Excited to announce the 1st Workshop on Physical AI: Understanding and Building the Physical World (PhysAI) at #ECCV2026 in Malmö, Sweden!
📅 Sept. 9, 2026
🌐 Website: phys-ai-workshop.github.io/e…
A bit of a backstory, Instruct-Particulate is the project that we made Articraft for.
It was very difficult to source clean 3D assets with kinematic data, so we thought about how to generate it synthetically.
🚀 Introducing Instruct-Particulate, our new model for inferring articulated structures from static 3D meshes, with significantly improved generalization to novel object categories and support for kinematic prompting.
To achieve this, we scaled our training data 40× and redesigned the model to follow kinematic prompts.
The result: diverse, realistic, simulator-compatible articulated 3D assets can now be generated directly from real-world images!
🔗 Project page: instruct-particulate.github.…
🤗 Demo: huggingface.co/spaces/rayli/…
While this looks like a nice-to-have feature that enables test-time prompting, this design is actually essential during training! There is often no single correct way to assign an articulated structure to a given 3D object. Annotation inconsistencies in part granularity and semantics would lead a naively trained model to "average" over multiple plausible annotations and produce suboptimal results. The kinematic context disambiguates the target output, enabling all 4 heterogeneous datasets to be included in training.
We then add kinematic context to the model input, including an explicit kinematic structure (a list of parts and their connectivity), text prompt of each part, joint types (i.e., prismatic or revolute), and optional 3D point prompts.
(4/4) Generic part-segmented 3D models. Generic 3D part segmentations rarely correspond to object articulation and motion. We mix them in the training data as a pre-training step for aligning text prompts with 3D parts.
(3/4) Synthetic 3D objects pseudo-labeled with kinematic parts. We build an annotation pipeline that identifies and segments 3D kinematic parts of synthetic 3D objects using vision-language models.
🚀 Introducing Articraft, a coding agent for articulated 3D asset creation.
Articraft writes code, executes it, receives validation feedback, and refines the result into simulation-ready 3D assets with parts, joints, and motion.
We’re also releasing Articraft-10K: 10,000+ articulated objects across 250 categories, unlocking large-scale interactive scenes for robotics simulation and physical AI.
🔗 Project page: articraft3d.github.io/
💻 Code: github.com/mattzh72/articraf…
While articulated 3D models are widely used in gaming and robotics simulation, they are scarce on the Internet due to the cost of designing and annotating them manually. In developing Instruct-Particulate, we curate a massive dataset consisting of the following components:
🚀 Introducing Instruct-Particulate, our new model for inferring articulated structures from static 3D meshes, with significantly improved generalization to novel object categories and support for kinematic prompting.
To achieve this, we scaled our training data 40× and redesigned the model to follow kinematic prompts.
The result: diverse, realistic, simulator-compatible articulated 3D assets can now be generated directly from real-world images!
🔗 Project page: instruct-particulate.github.…
🤗 Demo: huggingface.co/spaces/rayli/…
CAMEL-AI Live Talk this Friday 🚨
"Articraft: An Agentic System for Scalable Articulated 3D Asset Generation" by Matt Zhou @Mattzh1314 (Visiting Researcher at the University of Cambridge)
Articraft is a system for generating 3D articulated assets at scale. It leverages a custom agentic harness with geometric validation built in, and a LLM-friendly 3D design library to achieve this. Tuning for a blend of both realism and cost-effectiveness, Matt and his team use Articraft to create a large 10k scale dataset of articulated objects, and study using this data to improve existing models and unlock downstream applications in robotic simulation and VR.
Paper link: arxiv.org/html/2605.15187v1
🗓 May 29 · 8:00 PT / 16:00 BST
🔗 Register for Live Talk: forms.gle/K1drb9ao5QQmhB6z8