Document Question Answering on ArxivQA (test)
76.03AccuracyOracle
Evaluation Results
| Method | Links | |
|---|---|---|
| OracleModel=MiniCPM-V 2.62026.02 | 76.03 | |
| MARAModel=MiniCPM-V 2.62026.02 | 71.85 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-32026.02 | 70.07 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-102026.02 | 69.65 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-12026.02 | 69.47 | |
| OracleModel=GPT-4o2026.02 | 67.58 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-32026.02 | 65.76 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-12026.02 | 65.39 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-102026.02 | 64.78 | |
| MARAModel=GPT-4o2026.02 | 63.92 | |
| MARA (Our Framework)Model=GPT-4o, Input=top-32026.02 | 63.9 | |
| MARA (Our Framework)Model=GPT-4o, Input=top-12026.02 | 63.19 | |
| VisRAGModel=GPT-4o, Input=top-12026.02 | 62.94 | |
| VisRAGModel=GPT-4o, Input=top-32026.02 | 62.26 |