Chart Visual Question Answering on CHARTOM non-misleading visualizations
89.5QA AccuracyGPT5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT52025.02 | 89.5 | — | |
| GPT5-mini2025.02 | 81.5 | — | |
| claude-3.5-sonnet2025.02 | 71.8 | 63.7 | |
| Claude-3.5-sonnet2025.02 | 71.8 | — | |
| gemini-1.5-pro2025.02 | 66.9 | 58.1 | |
| Gemini-2.5-flash2025.02 | 66.9 | — | |
| gpt4o2025.02 | 64.5 | 55.7 | |
| internvl2.5-38BModel size=38B2025.02 | 63.7 | 54.8 | |
| gemini-1.5-flash2025.02 | 59.7 | 50.8 | |
| ovis1.6-9BModel size=9B2025.02 | 57.3 | 48.4 | |
| internvl2.5-8BModel size=8B2025.02 | 57.3 | 48.4 | |
| ovis1.6-27BModel size=27B2025.02 | 55.7 | 46.8 | |
| internvl2.5-26BModel size=26B2025.02 | 55.7 | 46.8 | |
| gpt42025.02 | 52.4 | 43.6 | |
| qwen2vl-7BModel size=7B2025.02 | 50.8 | 41.9 | |
| internvl2.5-4BModel size=4B2025.02 | 50 | 41.1 | |
| internvl2.5-2BModel size=2B2025.02 | 44.4 | 35.5 | |
| llava-7BModel size=7B2025.02 | 42.7 | 33.9 | |
| llava-13BModel size=13B2025.02 | 41.1 | 32.3 | |
| qwen2vl-2BModel size=2B2025.02 | 39.5 | 30.7 | |
| chartinstruction-13BModel size=13B2025.02 | 35.5 | 26.6 | |
| chartgemma-3BModel size=3B2025.02 | 23.4 | 16.1 | |
| tinychart-3BModel size=3B2025.02 | 22.6 | 15.3 |