Grounded 3D-Aware Spatial Vision-Language Modeling

Benchmark Model Rank Results
3d-object-detection-on-omni3dGR3D-8B–AP3D: 25.4
chart-question-answering-on-chartqaGR3D-8B–1:1 Accuracy: 81.01
object-hallucination-evaluation-on-popeGR3D-8B–F1: 88.23
spatial-reasoning-on-vsi-benchGR3D-8B–Accuracy: 67.6