Chart Question Answering on CDCC v1 (test)
79.09AccuracyChartCynics
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ChartCynicsBackbone=GPT-o4-mini, Configuration=Train-free2026.03 | 79.09 | 9.09 | 11.82 | |
| ChartCynicsBackbone=Gemini-2.5-Flash, Configuration=Train-free2026.03 | 75.45 | 7.27 | 17.27 | |
| GPT-o4-miniConfiguration=Base Model2026.03 | 69.09 | 17.27 | 13.64 | |
| Gemini-3.1-Procategory=Standard Multimodal Baseline2026.03 | 68.18 | 14.54 | 17.27 | |
| ChartCynicsBackbone=Qwen3-VL-8B, Configuration=SFT + GRPO (Full)2026.03 | 64.55 | 8.18 | 27.27 | |
| Gemini-2.5-FlashConfiguration=Base Model2026.03 | 62.73 | 17.27 | 20 | |
| ChartCynicsBackbone=Qwen3-VL-8B, Configuration=Train-free2026.03 | 47.47 | 16.16 | 36.36 | |
| RedrawnBackbone=Qwen3-VL-8B, category=Structural Mitigation Heuristic2026.03 | 38.18 | 20.91 | 40.91 | |
| Table-based QABackbone=Qwen3-VL-8B, category=Structural Mitigation Heuristic2026.03 | 36.39 | 20.91 | 42.73 | |
| Qwen3-VL-8BConfiguration=Base Model2026.03 | 35.45 | 31.82 | 32.73 | |
| ChartMoEcategory=Standard Multimodal Baseline2026.03 | 28.18 | 46.36 | 25.45 |