Measurement Reading on MeasureBench Real-world
30.2Overall AccuracyGemini-2.5-Pro
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini-2.5-ProSource Type=Closed-source2026.04 | 30.2 | 31.5 | 80.2 | 21.9 | 3.8 | |
| GPT-5-MiniSource Type=Closed-source2026.04 | 22 | 20.8 | 70.8 | 16.9 | 2.9 | |
| Qwen3-VL-4B-Instruct + TriSCASource Type=Open-source2026.04 | 20.5 | 28.5 | 41.7 | 10.4 | 1.9 | |
| Gemini-2.5-FlashSource Type=Closed-source2026.04 | 20.2 | 20.5 | 65.6 | 13 | 1 | |
| GPT-5Source Type=Closed-source2026.04 | 19.8 | 18.3 | 66.7 | 15.2 | 2.9 | |
| Qwen3-VL-8BSource Type=Open-source2026.04 | 15.3 | 14.5 | 53.1 | 11.3 | 0 | |
| Qwen2.5-VL-7BSource Type=Open-source2026.04 | 14.6 | 13.8 | 49 | 11.4 | 0 | |
| Qwen2.5-VL-72BSource Type=Open-source2026.04 | 14.5 | 12.2 | 55.2 | 12.2 | 0 | |
| Qwen3-VL-4B-InstructSource Type=Open-source2026.04 | 13.7 | 12 | 43.8 | 11.9 | 1 | |
| Qwen2.5-VL-32BSource Type=Open-source2026.04 | 11.7 | 9 | 51.6 | 9.7 | 0 |