Leaderboard
Scene generation models evaluated by GREP.
Wall-free single roomsExplore the examples
Overall ranking
0–100 · Higher is better
Small gaps do not establish a reliable ordering. The repeatability study estimates that differences below about 4 points can arise from one evaluation per scene. Appendix D.1 ↗
Loading results…
Select a model to view available examples. Table 1 ↗
Score comparison Overall scores at a glance
Avg. Overall
Scoring protocol Metric weights and aggregation
Loading scoring protocol…
Paper experimental setup ↗ · Task definitions and generation settings.