- Who it helps
- Researchers and developers evaluating generated interactive worlds.
- Availability
- The October 14 abstract reports an evaluation covering 720 instances and 27 models. Linked dataset and implementation availability were not independently inspected.
How it works
Its tests separate preserving input structure, maintaining new generated content and staying consistent with earlier views. A plausible single frame can miss those temporal failures.
What to check for your task
Look for a repeated camera path and revisit examples when reviewing a world-model demonstration. Compare the same objects before and after the return.
What you can do
- Look for a repeated camera path and revisit examples when reviewing a world-model demonstration. Compare the same objects before and after the return. Open reference ↗
Where the evidence stops
This review read the abstract and did not reproduce scores or inspect every environment. Results should stay tied to the reported protocol.
Sources read. Public sources read October 18, 2026. Product setup, inference and end-to-end behavior were not executed.
Sources & further reading
Paper · paper
Orbia research abstract ↗
Read the abstract’s spatial-consistency evaluation and reported sample. Full dataset and results were not reproduced.
Orbia authors · page checked 2026-10-18 · Published 2026-10-14
Coverage note: This bounded review lacks a complete cross-platform discussion sample. X search required sign-in; independent execution evidence remains unavailable.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
Orbia evaluates whether an interactive video system preserves a scene when the viewer explores it and returns.