Optical Character Recognition on Tibetan (test)
90.37AccuracyOmniOCR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OmniOCREvaluation Protocol=Dynamic LoRA adaptation2026.02 | 90.37 | 91.12 | 90.48 | |
| RolmOCREvaluation Protocol=Full Fine-tune2026.02 | 89.21 | 89.92 | 89.33 | |
| RolmOCREvaluation Protocol=Fine-tuning (LoRA)2026.02 | 80.52 | 79.84 | 80.58 | |
| Qwen-VL-OCREvaluation Protocol=Zero-Shot2026.02 | 38.85 | 37.96 | 38.18 | |
| Claude-3.7-SonnetEvaluation Protocol=Zero-Shot2026.02 | 34.63 | 33.87 | 34.08 | |
| Doubao-1.5-Vision-ProEvaluation Protocol=Zero-Shot2026.02 | 32.55 | 31.42 | 31.76 | |
| InternVL3-78BEvaluation Protocol=Zero-Shot2026.02 | 30.82 | 32.67 | 31.71 | |
| Pixtral LargeEvaluation Protocol=Zero-Shot2026.02 | 30.56 | 25.89 | 27.98 | |
| Kimi-VLEvaluation Protocol=Zero-Shot2026.02 | 29.45 | 28.73 | 28.96 | |
| RolmOCREvaluation Protocol=Zero-Shot2026.02 | 29.31 | 28.57 | 28.94 | |
| Moonshot-v1Evaluation Protocol=Zero-Shot2026.02 | 28.84 | 31.12 | 29.95 | |
| DeepSeek-VL2Evaluation Protocol=Zero-Shot2026.02 | 27.83 | 32.95 | 30.21 | |
| Gemini 2.5 ProEvaluation Protocol=Zero-Shot2026.02 | 27.41 | 26.73 | 26.96 | |
| GLM-4v-PlusEvaluation Protocol=Zero-Shot2026.02 | 27.24 | 26.43 | 26.71 | |
| Qwen-VL-MaxEvaluation Protocol=Zero-Shot2026.02 | 26.15 | 27.32 | 27.06 | |
| GPT-4oEvaluation Protocol=Zero-Shot2026.02 | 25.61 | 27.18 | 26.37 |