Treellis Wins Best Paper Award at UIST 2026

Our paper "Treellis: Exposing Editable Assembly Structures for Progressive Human-AI Co-Creation of 3D Models" has been accepted at ACM UIST 2026, the 39th Annual ACM Symposium on User Interface Software and Technology in Detroit, and has received a Best Paper Award! 🏆

The work was led by PhD student Yixiang Chen, together with Adele Tong, Praneeth Perera, Chuancheng Shi, Per Ola Kristensson (University of Cambridge), Zhiyong Wang, Rajesh Balan (Singapore Management University), Alistair McEwan, Zhanna Sarsenbayeva and Anusha Withana.

Beyond Prompt-Only 3D Generation

Generative AI makes 3D modeling more accessible, but prompt-only tools return a finished model with little visibility into how it is built. Fixing one part often means re-prompting and watching unrelated parts change.

Treellis takes a different approach. From a text prompt, it generates a 3D object together with its assembly structure:

• Structure Tree: shows how the object is assembled from parts, operations and repeats.
• Parameter Panel: exposes part-level sliders for position, rotation, scale, size and colour.
• Scoped editing: users can tweak parameters, restructure nodes by hand, or prompt the AI on just the selected part, leaving the rest of the model untouched.

Key Findings

In a controlled study (N=24) against a prompt-only baseline, Treellis raised independent task completion from 92/108 to 106/108 and improved perceived usability (SUS 87.7 vs. 74.0). A multi-session open-ended study (N=5) showed participants developing efficient hybrid strategies that mix AI generation with manual edits to build complex designs.

Try It

Treellis is open-source and runs in the browser. Try it at treellis.art, and find the paper, figures and BibTeX on the project page or via DOI.

Congratulations to Yixiang and the whole team!