| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| View Synthesis | ABO (test) | PSNR33.13 | 18 | |
| 3D Object Captioning | ABO 6.4k objects | CLIPScore82.3 | 9 | |
| Single-image 3D reconstruction | ABO dataset (test) | FID27.88 | 7 | |
| Normal Map Geometry Reconstruction | ABO | PSNR27.3252 | 7 | |
| Point Cloud Geometry Reconstruction | ABO | Chamfer Distance (L1)0.541 | 7 | |
| Appearance Reconstruction | ABO | PSNR25.14 | 7 | |
| Inpainting | ABO (test) | CLIP Score28.36 | 7 | |
| Text-to-Image Factuality Evaluation | ABO 50 | FAGER Score88.23 | 6 | |
| Part-level 3D object generation | ABO | r-FID4.5632 | 5 | |
| 3D Object Generation | ABO | CD0.101 | 5 | |
| Part-aware 3D Reconstruction | ABO | CD0.092 | 5 | |
| classification | ABO | Top-1 Acc61.7 | 5 | |
| Novel View Synthesis | ABO standard (OOD) | Prediction Error0.0065 | 4 | |
| Part-based 3D Object Generation | ABO | IoU4.5 | 4 | |
| Part-Composed 3D Object Generation | ABO (randomly sampled 100 objects) | Self-IoU0.0139 | 4 | |
| Text-3D Retrieval | ABO | Top-1 Accuracy15.87 | 4 | |
| Image-3D Retrieval | ABO | Top-1 Accuracy66.15 | 4 | |
| 3D Reconstruction | ABO (test) | PSNR30.92 | 4 | |
| 3D Reconstruction | ABO | PSNR29.09 | 4 | |
| Single-view NeRF Generation | ABO Sofa 512x512 (test) | PSNR23.96 | 4 | |
| single-view NeRF generation | ABO Sofa (test) | PSNR26.73 | 4 | |
| single-view NeRF generation | ABO Chairs (test) | PSNR25.92 | 4 | |
| Geometry Captioning | ABO Fine-Grained Geometry Captions (test) | Win %88.21 | 4 | |
| Factual A/B test | ABO 50 pairs (test) | Pairwise Accuracy82 | 3 | |
| Novel View Synthesis | ABO Day-to-Night standard (test) | Prediction Error0.0039 | 2 |