Driving with language reasoning on DriveLM (leaderboard)
83AccuracyHERMES++
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HERMES++Reference=-2026.04 | 83 | 0.61 | 43 | 0.59 | |
| HERMES++Reference=-2026.04 | 83 | 0.61 | 43 | 0.59 | |
| Team NVIDIAReference=CVPRW 24 [92]2026.04 | 78 | 0.6 | — | 0.59 | |
| Omni-QReference=CVPR 25 [15]2026.04 | 78 | 0.64 | 37 | 0.58 | |
| Team NVIDIAReference=CVPRW 242026.04 | 78 | 0.6 | — | 0.59 | |
| Omni-QReference=CVPR 252026.04 | 78 | 0.64 | 37 | 0.58 | |
| FSDriveReference=NeurIPS 25 [54]2026.04 | 72 | 0.63 | 39 | 0.57 | |
| FSDriveReference=NeurIPS 252026.04 | 72 | 0.63 | 39 | 0.57 | |
| MMFM ADReference=CVPRW 24 [92]2026.04 | 67 | 0.64 | — | 0.57 | |
| MMFM ADReference=CVPRW 242026.04 | 67 | 0.64 | — | 0.57 | |
| DriveLMReference=ECCV 24 [16]2026.04 | 65 | 0.53 | 28 | 0.5 | |
| DriveLMReference=ECCV 242026.04 | 65 | 0.53 | 28 | 0.5 |