Field-level Extraction on Private KYC Dataset
87.27Field-level AccuracyMiniCPM-o-2.6
Evaluation Results
| Method | Links | |
|---|---|---|
| MiniCPM-o-2.6OCR Engine=PaddleOCR, Pipeline Configuration=Full2026.04 | 87.27 | |
| Qwen3-VL-8B-InstructOCR Engine=PaddleOCR, Pipeline Configuration=Full2026.04 | 85.3 | |
| MiniCPM-o-2.6OCR Engine=PaddleOCR, Pipeline Configuration=-Prompt2026.04 | 84.38 | |
| Qwen3-VL-8B-InstructOCR Engine=PaddleOCR, Pipeline Configuration=-Prompt2026.04 | 80.97 | |
| Qwen3-VL-8B-InstructOCR Engine=PaddleOCR, Pipeline Configuration=-ImgPrep2026.04 | 77.17 | |
| MiniCPM-o-2.6OCR Engine=EasyOCR, Pipeline Configuration=Full2026.04 | 75.2 | |
| MiniCPM-o-2.6OCR Engine=EasyOCR, Pipeline Configuration=-Prompt2026.04 | 74.28 | |
| Gemma-3-27b-itOCR Engine=PaddleOCR, Pipeline Configuration=Full2026.04 | 72.97 | |
| Gemma-3-27b-itOCR Engine=PaddleOCR, Pipeline Configuration=-Prompt2026.04 | 72.57 | |
| MiniCPM-o-2.6OCR Engine=PaddleOCR, Pipeline Configuration=-ImgPrep2026.04 | 71 | |
| MiniCPM-o-2.6OCR Engine=EasyOCR, Pipeline Configuration=-ImgPrep2026.04 | 67.59 | |
| Gemma-3-27b-itOCR Engine=EasyOCR, Pipeline Configuration=Full2026.04 | 65.09 | |
| Gemma-3-27b-itOCR Engine=EasyOCR, Pipeline Configuration=-Prompt2026.04 | 64.3 | |
| Qwen3-VL-8B-InstructOCR Engine=PaddleOCR, Pipeline Configuration=-Retrieval2026.04 | 63.78 | |
| MiniCPM-o-2.6OCR Engine=PaddleOCR, Pipeline Configuration=-Retrieval2026.04 | 63.25 | |
| Gemma-3-27b-itOCR Engine=PaddleOCR, Pipeline Configuration=-ImgPrep2026.04 | 60.89 | |
| MiniCPM-o-2.6OCR Engine=EasyOCR, Pipeline Configuration=-Retrieval2026.04 | 59.45 | |
| Gemma-3-27b-itOCR Engine=EasyOCR, Pipeline Configuration=-ImgPrep2026.04 | 58.92 | |
| Qwen3-VL-8B-InstructOCR Engine=PaddleOCR, Pipeline Configuration=Direct VLM (Baseline)2026.04 | 55.91 | |
| MiniCPM-o-2.6OCR Engine=PaddleOCR, Pipeline Configuration=Direct VLM (Baseline)2026.04 | 55.38 | |
| MiniCPM-o-2.6OCR Engine=EasyOCR, Pipeline Configuration=Direct VLM (Baseline)2026.04 | 54.2 | |
| Gemma-3-27b-itOCR Engine=PaddleOCR, Pipeline Configuration=-Retrieval2026.04 | 52.1 | |
| Gemma-3-27b-itOCR Engine=EasyOCR, Pipeline Configuration=-Retrieval2026.04 | 48.29 | |
| Gemma-3-27b-itOCR Engine=PaddleOCR, Pipeline Configuration=Direct VLM (Baseline)2026.04 | 47.64 | |
| Gemma-3-27b-itOCR Engine=EasyOCR, Pipeline Configuration=Direct VLM (Baseline)2026.04 | 47.51 |