End-to-End Autonomous Driving on WOD-E2E (val)
57Latency (ms)RAP-DINO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RAP-DINOConfiguration=VA, ∼0.88B, Batch size=1, Hardware=1× H2002026.05 | 57 | 17.68 | — | 7.91 | |
| MindVLA-U1Configuration=vqa_first_fast, template QA, 1 step, Batch size=1, Hardware=1× H2002026.05 | 63 | 15.92 | 5,412 | 7.67 | |
| MindVLA-U1Configuration=InternVL-2 1B, template QA, 2 steps, Batch size=1, Hardware=1× H2002026.05 | 64 | 15.55 | 2,478 | 7.78 | |
| MindVLA-U1Configuration=action_only, 2 steps, Batch size=1, Hardware=1× H2002026.05 | 103 | 9.7 | 5,412 | 7.74 | |
| MindVLA-U1Configuration=vqa_first_fast, template QA, 2 steps, Batch size=1, Hardware=1× H2002026.05 | 108 | 9.26 | 5,412 | 7.82 | |
| MindVLA-U1Configuration=vqa_first_decoding, no intent, 2 steps, Batch size=1, Hardware=1× H2002026.05 | 2,594 | 0.39 | 5,431 | 7.83 |