Driving Reward Modeling on DriveReward Bench
80.6SafetyDriveReward
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DriveRewardParam=1B, Evaluation Protocol=Task-Specific2026.06 | 80.6 | 81.2 | 99.9 | 97.6 | 23 | |
| InternVL3-1BParam=1B, Evaluation Protocol=Task-Specific2026.06 | 72.2 | 75.6 | 95.5 | 92.2 | 34 | |
| InternVL3Param=8B, Evaluation Protocol=Zero-Shot2026.06 | 64.2 | 63.7 | 64.2 | 69.3 | 58 | |
| Qwen3Param=4B, Evaluation Protocol=Zero-Shot2026.06 | 60.5 | 63.9 | 63.8 | 59.3 | 55 | |
| Qwen3.5Param=4B, Evaluation Protocol=Zero-Shot2026.06 | 59.9 | 61.5 | 80.1 | 78.8 | 60 | |
| Qwen3Param=8B, Evaluation Protocol=Zero-Shot2026.06 | 58.8 | 62.4 | 71.4 | 73.8 | 55 | |
| Qwen3.5Param=2B, Evaluation Protocol=Zero-Shot2026.06 | 30.3 | 59.3 | 74.4 | 80.6 | 57 | |
| InternVL3Param=2B, Evaluation Protocol=Zero-Shot2026.06 | 29.6 | 61.4 | 67.2 | 67.5 | 62 |