Visual Question Answering on V2X-QA infrastructure-side 1.0 (test)
99.6IS1 AccuracyV2X-MoE-8B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| V2X-MoE-8BModel Type=Open-Source2026.04 | 99.6 | 95.1 | 92.1 | 89.1 | 94 | |
| Qwen3-VL-FlashModel Type=Closed-Source2026.04 | 93.8 | 50 | 27.8 | 47.2 | 54.7 | |
| Intern-S1Model Type=Open-Source2026.04 | 82.3 | 61.5 | 60.8 | 34.3 | 59.7 | |
| Qwen-2.5-VL-72B-InstructModel Type=Open-Source2026.04 | 71.7 | 58.1 | 63 | 46 | 59.7 | |
| GPT-5.2Model Type=Closed-Source2026.04 | 69.8 | 42.6 | 38.5 | 47.5 | 49.6 | |
| Gemini-3-Flash-PreviewModel Type=Closed-Source2026.04 | 63.8 | 44.9 | 20.8 | 11.7 | 35.3 | |
| Qwen-3-VL-30B-A3B-InstructModel Type=Open-Source2026.04 | 63.8 | 45.7 | 40.8 | 53.2 | 50.9 | |
| Intern-S1-MiniModel Type=Open-Source2026.04 | 55.8 | 55.1 | 41.9 | 41.1 | 48.5 | |
| Qwen-3.5-PlusModel Type=Closed-Source2026.04 | 53.2 | 56.6 | 44.5 | 39.6 | 48.5 | |
| GPT-5-MiniModel Type=Closed-Source2026.04 | 45.7 | 52.1 | 46.4 | 45.3 | 47.4 | |
| Gemini-2.5-ProModel Type=Closed-Source2026.04 | 24.5 | 43 | 45.3 | 51.7 | 41.1 |