Text-to-Code on VisPlotMermaid
90.1Execution Success RateGemini-3-Flash
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-3-Flashtype=Closed-Source2026.04 | 90.1 | 58.4 | 80.2 | |
| OmniDia-3Bmode=RL, parameters=3B2026.04 | 88.6 | 49.4 | 64.5 | |
| OmniDia-7Bmode=RL, parameters=7B2026.04 | 86.3 | 51 | 66.9 | |
| OmniDia-7Bmode=SFT, parameters=7B2026.04 | 84 | 45.3 | 62.6 | |
| OmniDia-3Bmode=SFT, parameters=3B2026.04 | 82.4 | 45.8 | 64.3 | |
| Qwen3-VL-32Bbackbone=Qwen3-VL, parameters=32B2026.04 | 67.2 | 40.4 | 55.1 | |
| InternVL3.5-8Bbackbone=InternVL3.5, parameters=8B2026.04 | 64.9 | 34.7 | 48.4 | |
| InternVL3.5-14Bbackbone=InternVL3.5, parameters=14B2026.04 | 63.4 | 35.2 | 49 | |
| InternVL3.5-38Bbackbone=InternVL3.5, parameters=38B2026.04 | 63.4 | 33.8 | 45.8 | |
| Qwen2.5-VL-72Bbackbone=Qwen2.5-VL, parameters=72B2026.04 | 60.3 | 31 | 46 | |
| GPT-5-minitype=Closed-Source2026.04 | 57.5 | 34.2 | 50.5 | |
| Qwen2.5-VL-7Bbackbone=Qwen2.5-VL, parameters=7B2026.04 | 57.3 | 25.6 | 37.3 | |
| Qwen2.5-VL-3Bbackbone=Qwen2.5-VL, parameters=3B2026.04 | 51.2 | 22.4 | 31.9 | |
| InternVL3-8Bbackbone=InternVL3, parameters=8B2026.04 | 45 | 19.1 | 28.2 |