Chart Question Answering on ChartMuseum (OA, RA, VA, RVC)
73.2Overall Accuracy (OA)Gemini 2.5 Pro
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini 2.5 ProModel Group=Proprietary models2026.05 | 73.2 | 70.4 | 52.9 | -27.8 | |
| GPT-5.4Model Group=Proprietary models2026.05 | 70.4 | 71.8 | 55.4 | -21.3 | |
| Claude Sonnet 4.6Model Group=Proprietary models2026.05 | 69 | 70.4 | 54.1 | -21.6 | |
| Gemini 2.5 FlashModel Group=Proprietary models2026.05 | 64.1 | 60.6 | 49.3 | -23.1 | |
| GPT-4.1 miniModel Group=Proprietary models2026.05 | 52.8 | 52.8 | 44.6 | -15.5 | |
| GPT-4.1Model Group=Proprietary models2026.05 | 50.7 | 54.9 | 39.2 | -22.6 | |
| GPT-5.4 miniModel Group=Proprietary models2026.05 | 48.6 | 51.4 | 42.5 | -12.5 | |
| Claude Haiku 4.5Model Group=Proprietary models2026.05 | 47.2 | 45.1 | 35.7 | -24.3 | |
| Qwen3-VL-8B-InstructModel Group=Open-source models2026.05 | 38 | 42.3 | 33.9 | -10.7 | |
| GPT-4oModel Group=Proprietary models2026.05 | 34.5 | 36.6 | 28.7 | -16.7 | |
| Gemma 4 E4B ITModel Group=Open-source models2026.05 | 28.9 | 26.1 | 25.1 | -12.9 | |
| Qwen2.5-VL-7B-InstructModel Group=Open-source models2026.05 | 27.5 | 21.1 | 22.3 | -18.7 | |
| InternVL3-8BModel Group=Open-source models2026.05 | 26.8 | 23.9 | 22.3 | -16.6 | |
| Pixtral-12BModel Group=Open-source models2026.05 | 24.6 | 25.4 | 21.1 | -14.6 | |
| LLaVA-OneVision-Qwen2-7BModel Group=Open-source models2026.05 | 19 | 21.1 | 16.4 | -13.7 |