View-Conditioned Spatial Reasoning on ViewBench Shape
100AccuracyTarget View (Oracle)
Evaluation Results
| Method | Links | |
|---|---|---|
| Target View (Oracle)View Overlap (%)=5–15, Depth=GT2026.04 | 100 | |
| Target View (Oracle)View Overlap (%)=15–25, Depth=GT2026.04 | 100 | |
| Target View (Oracle)View Overlap (%)=25–35, Depth=GT2026.04 | 100 | |
| Backward-Adaptive Token WarpingView Overlap (%)=25–35, Depth=GT2026.04 | 75.72 | |
| Backward-Nearest Token WarpingView Overlap (%)=25–35, Depth=Pred.2026.04 | 75.31 | |
| Backward-Nearest Token WarpingView Overlap (%)=25–35, Depth=GT2026.04 | 73.25 | |
| Backward-Adaptive Token WarpingView Overlap (%)=25–35, Depth=Pred.2026.04 | 70.37 | |
| Backward-Nearest Token WarpingView Overlap (%)=5–15, Depth=GT2026.04 | 67.44 | |
| Backward-Adaptive Token WarpingView Overlap (%)=5–15, Depth=GT2026.04 | 67.44 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=25–35, Depth=GT2026.04 | 66.67 | |
| Backward-Adaptive Token WarpingView Overlap (%)=5–15, Depth=Pred.2026.04 | 66.28 | |
| Backward-Adaptive Token WarpingView Overlap (%)=15–25, Depth=GT2026.04 | 66.26 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=25–35, Depth=Pred.2026.04 | 64.2 | |
| Backward-Nearest Token WarpingView Overlap (%)=5–15, Depth=Pred.2026.04 | 63.95 | |
| Backward-Nearest Token WarpingView Overlap (%)=15–25, Depth=GT2026.04 | 62.96 | |
| Backward-Nearest Token WarpingView Overlap (%)=15–25, Depth=Pred.2026.04 | 62.55 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=5–15, Depth=GT2026.04 | 62.4 | |
| Backward-Adaptive Token WarpingView Overlap (%)=15–25, Depth=Pred.2026.04 | 61.32 | |
| Forward Token WarpingView Overlap (%)=15–25, Depth=Pred.2026.04 | 60.91 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=25–35, Depth=GT2026.04 | 60.49 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=25–35, Depth=Pred.2026.04 | 60.08 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=5–15, Depth=Pred.2026.04 | 58.14 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=15–25, Depth=GT2026.04 | 58.02 | |
| Forward Token WarpingView Overlap (%)=5–15, Depth=Pred.2026.04 | 56.98 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=15–25, Depth=GT2026.04 | 56.79 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=15–25, Depth=Pred.2026.04 | 56.79 | |
| Pixel-Wise Warping (Backward)View Overlap (%)=15–25, Depth=Pred.2026.04 | 56.79 | |
| Forward Token WarpingView Overlap (%)=25–35, Depth=Pred.2026.04 | 56.38 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=5–15, Depth=GT2026.04 | 56.2 | |
| Pixel-Wise Warping (Forward)View Overlap (%)=5–15, Depth=Pred.2026.04 | 56.2 | |
| GenWarpView Overlap (%)=25–35, Depth=GT2026.04 | 55.14 | |
| Forward Token WarpingView Overlap (%)=15–25, Depth=GT2026.04 | 55.14 | |
| Forward Token WarpingView Overlap (%)=5–15, Depth=GT2026.04 | 55.04 | |
| GenWarpView Overlap (%)=5–15, Depth=GT2026.04 | 53.1 | |
| Forward Token WarpingView Overlap (%)=25–35, Depth=GT2026.04 | 53.09 | |
| VLM-3RView Overlap (%)=25–35, Depth=GT2026.04 | 50.21 | |
| VLM-3RView Overlap (%)=15–25, Depth=GT2026.04 | 49.79 | |
| VLM-3RView Overlap (%)=5–15, Depth=GT2026.04 | 49.22 | |
| SpatialReasonerView Overlap (%)=25–35, Depth=GT2026.04 | 48.15 | |
| GenWarpView Overlap (%)=15–25, Depth=GT2026.04 | 47.33 | |
| SpatialReasonerView Overlap (%)=15–25, Depth=GT2026.04 | 38.27 | |
| Qwen2.5-VLView Overlap (%)=25–35, Depth=GT2026.04 | 37.86 | |
| ViLaSRView Overlap (%)=25–35, Depth=GT2026.04 | 34.57 | |
| SpatialReasonerView Overlap (%)=5–15, Depth=GT2026.04 | 33.72 | |
| Qwen2.5-VLView Overlap (%)=15–25, Depth=GT2026.04 | 25.1 | |
| Qwen2.5-VLView Overlap (%)=5–15, Depth=GT2026.04 | 24.42 | |
| ViLaSRView Overlap (%)=15–25, Depth=GT2026.04 | 23.05 | |
| ViLaSRView Overlap (%)=5–15, Depth=GT2026.04 | 22.87 |