Driving Reasoning on nuScenes (test)
52.96First-Frame Joint AccuracyDriveAgent-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DriveAgent-R1Tools=w/ Tools2025.07 | 52.96 | 47.1 | |
| DriveAgent-R1Tools=w/o Tools2025.07 | 52.58 | 44.43 | |
| HumanTools=w/o Tools2025.07 | 50.48 | 48.24 | |
| GPT-5Tools=w/ Tools2025.07 | 49.11 | 45.14 | |
| GPT-5Tools=w/o Tools2025.07 | 48.75 | 44.85 | |
| GPT-4.1Tools=w/ Tools2025.07 | 48.25 | 44.72 | |
| GPT-4.1Tools=w/o Tools2025.07 | 46.84 | 43.63 | |
| Gemini-2.5-FlashTools=w/ Tools2025.07 | 46.6 | 44.07 | |
| Doubao-Seed-1.6Tools=w/ Tools2025.07 | 46.29 | 43.1 | |
| Doubao-Seed-1.6Tools=w/o Tools2025.07 | 45.63 | 41.91 | |
| Gemini-2.5-FlashTools=w/o Tools2025.07 | 45.46 | 42.69 | |
| Qwen2.5-VL-72BTools=w/ Tools2025.07 | 43.87 | 40.47 | |
| Qwen2.5-VL-72BTools=w/o Tools2025.07 | 43.26 | 39.13 | |
| Qwen2.5-VL-7BTools=w/o Tools2025.07 | 33.11 | 27.6 | |
| Qwen2.5-VL-7BTools=w/ Tools2025.07 | 32.84 | 28.82 | |
| Qwen2.5-VL-3BTools=w/o Tools2025.07 | 30.18 | 23.48 | |
| Qwen2.5-VL-3BTools=w/ Tools2025.07 | 28.17 | 21.58 |