T3Bench
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Text-to-3D Generation | T3Bench (test) | Single Object Score58.3 | 14 | |
| Text-to-3D generation | T3Bench frozen (300-prompt audit set) | CLIP Score25.09 | 10 | |
| Measured orbit coverage | T3Bench 300-prompt derived (frozen audit set) | Metric A (Coverage Count/Score)359 | 9 | |
| Text-to-3D Human-Object Interaction Generation | T3Bench 100 prompts | CLIP Score31.4 | 7 | |
| Text-to-3D Generation | T3Bench 10 categories | Average T3Bench Score50.88 | 7 | |
| Text-to-3D Generation | T3Bench Object-centric | Imaging Score58.83 | 7 | |
| Appearance Consistency | T3Bench audit set (300 prompts) | Anchor-DINO Q10 Score56.8 | 4 | |
| Text-to-3D Generation | T3Bench prompt_scene | Quality Score18 | 1 |
Showing 8 of 8 rows