The authors introduce Artisan, built on top of the earlier Articraft approach, which used a single LLM agent to write and check 3D construction programs.
Multi-agent pipeline with simulator feedback produces physically usable 3D assets
Artisan separates design, coding, and validation across independent agents, lowering help collisions and improving stability across a new benchmark of more than 10 models.
Big Tech
Xudong Cai · Ruijie Zhu · Haoliang Shang · Ruining Li · Marc Pollefeys · Shangzhe Wu · +2 more
Nanyang Technological University · ETH Zurich · Microsoft · University of Cambridge · University of Oxford
Research Digest··3 min read
Artisan generates articulated 3D objects from text or reference images by delegating design, coding, and validation to independent agents, with feedback from a physics simulator.
Why this paper
From Microsoft and 4 others
In one line
A multi-agent pipeline using simulator feedback generates articulated 3D assets with competitive spatial fidelity and better collision avoidance and gravitational stability than evaluated alternatives.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§