GTA
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Multimodal tool-use | GTA | Answer Accuracy67.66 | 33 | |
| Semantic Segmentation | GTA to {Cityscapes, BDD100K, Mapillary, ACDC, DarkZurich} (val) | mIoU (Cityscapes)58.63 | 31 | |
| Trajectory Prediction | GTA-1M (test) | Path Error (Traj)626 | 17 | |
| Agent Task | GTA | Success Rate24.89 | 16 | |
| Tool Planning | GTA | Tool Selection F1 (Step-by-step)79.26 | 16 | |
| Semantic Segmentation | GTA to UAVID | Road IoU80.5 | 15 | |
| File Understanding and Generation | GTA | Score77.9 | 12 | |
| Step-level correctness prediction | GTA | AUROC92.17 | 10 | |
| Image-to-Image Translation | GTA to Cityscapes (test) | SSIM0.87 | 10 | |
| Visual Tool-use | GTA 121-case (val) | Tool Accuracy86.4 | 9 | |
| Image-to-Image Translation | GTA to KITTI (test) | SSIM0.82 | 9 | |
| Semantic Segmentation | GTA | Pixel Accuracy84.7 | 4 | |
| Autonomous LLM Fine-tuning | GTA | Accuracy72.2 | 4 | |
| Multi-agent recommendation | GTA | Top-1 Acc100 | 4 | |
| Single-agent tool selection | GTA | Top-1 Acc100 | 4 | |
| Inverse Dynamics Modeling | GTA V | Pearson Correlation X79.44 | 2 |
Showing 16 of 16 rows