Visually-rich Information Extraction on SROIE (test)
98.05ANLSQwen2-VL-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2-VL-7BTraining Data=RIDGE + RIDGE-DS2025.04 | 98.05 | |
| Qwen2-VL-7BTraining Data=RIDGE2025.04 | 97.74 | |
| Qwen2-VL-7BTraining Data=Original2025.04 | 97.5 | |
| InternVL2-8B-CoBDoc-CoB=true2025.05 | 97.181 | |
| DocOwl 1.5-CoBDoc-CoB=true2025.05 | 96.522 | |
| DocOwl 1.5-RReasoning tasks=true2025.05 | 96.35 | |
| Qwen2.5-VL-CoBDoc-CoB=true2025.05 | 96.07 | |
| InternVL2-2B-CoBDoc-CoB=true2025.05 | 95.94 | |
| DocOwl 1.5*SFT baseline=true2025.05 | 95.67 | |
| Qwen2.5-VL-RReasoning tasks=true2025.05 | 95.52 | |
| Qwen2.5-VL*SFT baseline=true2025.05 | 94.99 | |
| InternVL2-8B-RReasoning tasks=true2025.05 | 94.31 | |
| InternVL2-40BScale=40B2025.05 | 92.03 | |
| GPT-4o2025.05 | 91.87 | |
| InternVL2-8B*SFT baseline=true2025.05 | 90.39 | |
| Visual CoT2025.05 | 88.78 | |
| InternVL2-2B-RReasoning tasks=true2025.05 | 88.02 | |
| InternVL2-2B*SFT baseline=true2025.05 | 87.11 | |
| TextCoT2025.05 | 81.62 |