Visual Question Answering on OmniDrive (test)
86.67CI-rCogDriver-Agent
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CogDriver-AgentModel Scale=3B2025.08 | 86.67 | 36.4 | 7.52 | 47.58 | 27.87 | |
| LLava-nextModel Scale=72B2025.08 | 84.14 | 33.98 | 5.28 | 42.44 | 23.87 | |
| Qwen2.5VLModel Scale=72B2025.08 | 80.45 | 32.96 | 4.46 | 40.41 | 23.56 | |
| Qwen2.5VLModel Scale=7B2025.08 | 76.93 | 32.46 | 3.95 | 44.46 | 23.11 | |
| Qwen2.5VLModel Scale=32B2025.08 | 76.13 | 25.86 | 2.82 | 29.36 | 20.77 | |
| Qwen2.5VLModel Scale=3B2025.08 | 73.41 | 29.59 | 3.5 | 46.7 | 23.48 |