Real-time collaborative editing has a well-known failure mode: when two people work on the same thing at the same time, they get in each other’s way. Software solved this decades ago by not doing it — you branch, you work alone, you merge, and the tooling shows you where your changes collide.
Spatial collaboration never got that. Shared MR sessions are overwhelmingly synchronous and co-present in the same virtual space: everyone in the same model, editing live, watching each other’s hands.
CoBranchMR, posted September 23, 2026, imports the software workflow instead.
What it does
The system lets distributed collaborators work in parallel, from different locations, on the same digital representation of a physical object. The workflow:
- Branch the object into editable virtual copies
- Customise them independently
- Merge the work back into a shared object
- On merge, the system displays potential conflicts on the object’s surface and offers several resolution options
The authors are Niloofar Sayadi, Kaiyuan Tang, Yunhao Xing, Simret Gebreegziabher, Chaoli Wang and Diego Gomez-Zara.
Background, not this paper: a conventional shared MR workspace, showing the synchronous co-present model that CoBranchMR is proposing an alternative to.
The good idea is where the conflicts are shown
Branching a 3D model is not novel — every version control system for assets does something like it. The contribution is displaying conflicts on the object’s surface.
A merge conflict in text is legible because text is linear: two versions of line 47, shown side by side. A merge conflict in geometry is not. If two people both modified the same region of a shape, a textual or list-based conflict report (“Mesh_02 modified in both branches”) tells you almost nothing about what actually disagrees or whether it matters.
Painting the disagreement onto the geometry makes the conflict spatially legible. You walk around the object and see where the two versions diverge. That’s using the medium for what it’s actually good at — spatial understanding — rather than porting a text-based UI into 3D and calling it done.
Why this matters beyond research
The design of collaboration tools encodes an assumption about how creative work happens, and the synchronous-shared-space model encodes a bad one.
Everyone in the same room at the same time is a meeting. Meetings are where consensus gets manufactured and where the loudest person’s idea wins by default. Most good design work happens when people go away, try something properly, come back with a position, and argue about the actual artifacts.
The paper names this directly: the branch-and-merge model supports “parallel design, conflict resolution, and negotiation in remote creative work.” Negotiation requires that both parties have something concrete to negotiate over — two finished versions, not two half-formed suggestions made while someone else was talking.
For distributed studios, fabrication teams and anyone doing spatial design across time zones, an asynchronous workflow isn’t a compromise on real-time collaboration. It’s usually the better shape for the work, and the tooling has been the thing missing.
Related Reading
- CoBranchMR: Supporting Parallel Design and Conflict Resolution in Mixed Reality — arXiv:2609.27235
- Shared Mixed Reality Workspace — Photon Engine (YouTube)
- A Survey on Synchronous Augmented, Virtual and Mixed Reality Remote Collaboration Systems — arXiv
- CoCreatAR: Enhancing Authoring of Outdoor Augmented Reality Experiences Through Asymmetric Collaboration — arXiv
- arXiv cs.HC — Human-Computer Interaction listings