List your company

OOD Generalization on Real-world Cross-View Spatial Reasoning Benchmarks (Aggregate)

40Avg OOD PerformancePanoramic + VDrop
26.27229.83633.436.964May 26, 2026

Evaluation results

MethodLinks
Panoramic + VDropVDrop=✓2026.0540
Top-down + VDropVDrop=✓2026.0538
PanoramicVDrop=✗2026.0537.6
Top-downVDrop=✗2026.0537.3
ThinkMorph (24K)VDrop=✗2026.0537.2
Qwen3-VL-8BVDrop=✗2026.0537
Point Matching + VDropVDrop=✓2026.0536.1
Point MatchingVDrop=✗2026.0535.2
No-ThinkVDrop=✗2026.0535.1
Qwen3-VL-4BVDrop=✗2026.0533.8
BAGELVDrop=✗2026.0533.3
Text CoTVDrop=✗2026.0532.7
BAGEL-Zebra-CoT (182K)VDrop=✗2026.0526.8