Document Visual Question Answering on MP-DocVQA
89.5AccuracyGPT-4o-200b-128
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4o-200b-128Retrieval=Ground-Truth2025.10 | 89.5 | |
| InternVL2-8BRetrieval=Ground-Truth2025.10 | 84.8 | |
| M3DocRAGSize=10B, Paradigm=RAG2025.12 | 84.4 | |
| Pixtral-12B-2409Retrieval=Ground-Truth2025.10 | 83.9 | |
| InternVL3-8BSize=8B, Paradigm=End-to-End2025.12 | 83.3 | |
| GPT-4o-200b-128Retrieval=topk-adjusted2025.10 | 83.3 | |
| GPT-4o-200b-128Retrieval=LAD-RAG2025.10 | 82.9 | |
| GPT-4o-200b-128Retrieval=retrieving@52025.10 | 82.5 | |
| GPT-4o-200b-128Retrieval=retrieving@102025.10 | 81.9 | |
| DocopilotSize=8B, Paradigm=End-to-End2025.12 | 81.3 | |
| Phi-3.5-Vision-4BRetrieval=Ground-Truth2025.10 | 81 | |
| InternVL2-8BRetrieval=LAD-RAG2025.10 | 79.2 | |
| FRAGSize=8B, Paradigm=RAG2025.12 | 77.8 | |
| InternVL2-8BRetrieval=topk-adjusted2025.10 | 77.3 | |
| InternVL2-8BRetrieval=retrieving@102025.10 | 75.9 | |
| Phi-3.5-Vision-4BRetrieval=LAD-RAG2025.10 | 75.4 | |
| CogDocSize=7B, Paradigm=Unified Thinking2025.12 | 75 | |
| InternVL2-8BRetrieval=retrieving@52025.10 | 73.7 | |
| Phi-3.5-Vision-4BRetrieval=retrieving@102025.10 | 73.7 | |
| Phi-3.5-Vision-4BRetrieval=topk-adjusted2025.10 | 73.7 | |
| Phi-3.5-Vision-4BRetrieval=retrieving@52025.10 | 71.9 | |
| SV-RAGSize=8B, Paradigm=RAG2025.12 | 71 | |
| Pixtral-12B-2409Retrieval=LAD-RAG2025.10 | 70.4 | |
| mPLUG-DocOwl2Size=8B, Paradigm=End-to-End2025.12 | 69.4 | |
| Pixtral-12B-2409Retrieval=topk-adjusted2025.10 | 67.5 | |
| Pixtral-12B-2409Retrieval=retrieving@52025.10 | 66.3 | |
| Pixtral-12B-2409Retrieval=retrieving@102025.10 | 65.5 | |
| Qwen2.5-VL-72B-InstructSize=72B, Paradigm=End-to-End2025.12 | 64.2 | |
| MiMo-VL-7B-SFTSize=7B, Paradigm=End-to-End2025.12 | 63.9 | |
| Qwen2.5-VL-7B-InstructSize=7B, Paradigm=End-to-End2025.12 | 57.3 | |
| MiniCPM-Llama3-V2.5-8BRetrieval=All-Pages2025.10 | 19 | |
| Idefics2-8BRetrieval=All-Pages2025.10 | 17 | |
| mPLUG-DocOwl 1.5-8BRetrieval=All-Pages2025.10 | 15 |