Scene and Object Comprehension on Driving Evaluation Benchmark
89.3Small Object AccuracyUniUGP
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| UniUGPfull model=true2025.12 | 89.3 | 88.6 | 95.8 | |
| UniUGP (w/o CoT)Chain-of-Thought module=excluded2025.12 | 86.5 | 85.7 | 93.2 | |
| UniUGP (w/o Gen.)generation module=excluded2025.12 | 83.7 | 82.9 | 90.6 | |
| Qwen-2.5-VL-72Bhistorical trajectory information=provided, trajectory explanations=provided2025.12 | 75.8 | 74.9 | 81.5 | |
| GPT-4ohistorical trajectory information=provided, trajectory explanations=provided2025.12 | 64.2 | 63.5 | 72.8 |