Driving Explanation Generation on Nu-X
76.96BLEUFleetAgent
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FleetAgentInput Modality=Vector Embeddings2026.06 | 76.96 | 19.51 | 26.72 | |
| Gemini-2.5-FlashInput Modality=BEV Images2026.06 | 31.56 | 18.63 | 13.94 | |
| Gemini-2.5-FlashInput Modality=Language Description2026.06 | 27.77 | 18.67 | 12.75 | |
| Qwen-2.5VL-7BInput Modality=BEV Images2026.06 | 23.41 | 18.25 | 10.97 | |
| GPT-4oInput Modality=Language Description2026.06 | 14.35 | 11.93 | 5.85 | |
| ALN-P3*2026.06 | 5.59 | 14.7 | 35.2 | |
| HintAD*2026.06 | 4.18 | 13.2 | 27.6 | |
| GPT-4oInput Modality=Raw Images*2026.06 | 3.95 | 10.3 | 24.9 | |
| TOD3Cap*2026.06 | 2.45 | 10.5 | 23 | |
| Qwen-2.5VL-7BInput Modality=Language Description2026.06 | 2.37 | 8.28 | 4.29 |