Chart Understanding on ChartBench
79.33NQAGPT-5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5Model Size=-, # Training Data=-2026.05 | 79.33 | — | — | |
| Qwen2.5-VL + ChartCF (Ours)Model Size=7B, # Training Data=4K pairs (DPO)2026.05 | 72.43 | 82.45 | 81.34 | |
| Qwen3-VLModel Size=8B, # Training Data=-2026.05 | 72.33 | 71.43 | 71.53 | |
| Qwen3-VL + ECDModel Size=8B, # Training Data=300K (SFT)2026.05 | 71.95 | 81.42 | 80.37 | |
| Qwen3-VL + ChartCF (Ours)Model Size=8B, # Training Data=4K pairs (DPO)2026.05 | 71.95 | 83.36 | 82.09 | |
| Gemini-2.5-ProModel Size=-, # Training Data=-2026.05 | 71.24 | — | — | |
| Qwen2.5-VL + ECDModel Size=7B, # Training Data=300K (SFT)2026.05 | 70.86 | 79.35 | 78.41 | |
| InternVL3.5 + ChartCF (Ours)Model Size=8B, # Training Data=4K pairs (DPO)2026.05 | 67.95 | 79.87 | 78.55 | |
| Qwen2.5-VLModel Size=7B, # Training Data=-2026.05 | 67.81 | 80.99 | 79.53 | |
| InternVL3.5Model Size=8B, # Training Data=-2026.05 | 67.71 | 79.42 | 78.12 | |
| InternVL3.5 + ECDModel Size=8B, # Training Data=300K (SFT)2026.05 | 67.33 | 79 | 77.7 | |
| Qwen2.5-VL-72BParam.=72B2026.04 | 59.62 | 84.45 | — | |
| CHARTOOL-7BParam.=7B2026.04 | 58.95 | 91.58 | — | |
| Qwen3-VL-8BParam.=8B2026.04 | 57.86 | 76.83 | — | |
| Deepeyes-V2Param.=7B2026.04 | 57.76 | 71.05 | — | |
| Qwen2.5-VL-7BParam.=7B2026.04 | 57.67 | 87.58 | — | |
| DeepeyesParam.=7B2026.04 | 55.95 | 87.63 | — | |
| ECD Qwen2.5-VL-7BParam.=7B2026.04 | 53.9 | 75.52 | — | |
| ThymeParam.=7B2026.04 | 53.1 | 90.52 | — | |
| GPT-4oParam.=–2026.04 | 52.88 | 81.03 | — | |
| GPT-4oModel Size=-, # Training Data=-2026.05 | 52.88 | 81.03 | 77.9 | |
| CHARTOOL-3BParam.=3B2026.04 | 51.71 | 84.51 | — | |
| Claude Sonnet 3.5Param.=–2026.04 | 48.29 | 76.72 | — | |
| Claude-3.5-SonnetModel Size=-, # Training Data=-2026.05 | 48.29 | 76.72 | 73.56 | |
| Qwen2.5-VL-3BParam.=3B2026.04 | 48.1 | 77.26 | — | |
| ReachQA (InternVL2-8B)Model Size=8B, # Training Data=20K (SFT)2026.05 | 47.29 | 65.9 | 63.83 | |
| InternVL3-8BParam.=8B2026.04 | 45.95 | 84.18 | — | |
| GPT-4o miniModel Size=-, # Training Data=-2026.05 | 34.93 | 70.26 | 74.33 | |
| ChartGemmaModel Size=3B, # Training Data=123K (SFT)2026.05 | 34.1 | 78.9 | 73.92 | |
| ChartMOEParam.=8B2026.04 | 31.48 | 62.86 | — | |
| ChartGemmaParam.=3B2026.04 | 29.86 | 85.94 | — | |
| TinyChartParam.=3B2026.04 | 29.67 | 47.87 | — | |
| ChartReasonerModel Size=7B, # Training Data=140K (SFT)2026.05 | — | — | 55.2 |