Visual Question Answering on OmniEarth-Bench Open-ended question protocol 1.0 (test)
31.48Cross-sphere AccuracyGemini-2.0
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Gemini-2.0Model Type=Closed-source MLLM, Evaluation protocol=Open-ended2026.05 | 31.48 | 38.1 | 41.67 | 24.97 | 61.49 | 27.33 | 31.85 | 36.7 | |
| InternVL3-72BModel Type=Open-source MLLM, Evaluation protocol=Open-ended2026.05 | 29.51 | 39.14 | 27.51 | 32.45 | 53.87 | 38.29 | 34.67 | 36.49 | |
| GPT-4oModel Type=Closed-source MLLM, Evaluation protocol=Open-ended2026.05 | 25.76 | 23.21 | 33.13 | 25.17 | 46.46 | 13.65 | 17.17 | 26.36 | |
| Qwen2.5-VL-72BModel Type=Open-source MLLM, Evaluation protocol=Open-ended2026.05 | 24.78 | 22.08 | 38.62 | 31.17 | 15.23 | 20.22 | 16.87 | 24.14 |