Document Information Extraction on 90 Indian Document samples 1.0 (test)
100Aadhaar EM AccuracyParichay-1
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Parichay-1Rotation=Yes2026.02 | 100 | 60 | 90 | 77.5 | 85 | 75 | 100 | 72.22 | 74 | 81.42 | |
| Gemini 2.52026.02 | 100 | 80 | 85 | 62.5 | 90 | 65 | 94.44 | 66.67 | 69.5 | 79.2 | |
| Llama-4Parameters=17B2026.02 | 98 | 70 | 55 | 50.83 | 85 | 57.5 | 94.44 | 72.22 | 63.5 | 71.57 | |
| Phi4Rotation=Yes2026.02 | 93.33 | 65 | 90 | 55.83 | 83.33 | 77.5 | 90 | 60 | 65.71 | 75.04 | |
| Parichay-1Rotation=No2026.02 | 90 | 60 | 65 | 77.5 | 70 | 70 | 100 | 72.22 | 59.5 | 73.52 | |
| Parichay-1Rotation=Yes, Inference Engine=vLLM2026.02 | 90 | 15 | 75 | 40 | 90 | 50 | 100 | 38 | 60 | 62 | |
| Nanonets OCR-s2026.02 | 88 | 40 | 55 | 45.83 | 85 | 22.5 | 90 | 25 | 63 | 57.15 | |
| Gemma-3Parameters=27B, Training=IT2026.02 | 84 | 25 | 65 | 43.33 | 75 | 42.5 | 88.89 | 27.78 | 70 | 57.94 | |
| Mistral7BOCR=Azure2026.02 | 80 | 75 | 80 | 58.33 | 60 | 35 | 80 | 30 | 51.5 | 61.09 | |
| Mistral7BOCR=DocTR2026.02 | 52 | 35 | 40 | 50.83 | 35 | 27.5 | 45 | 25 | 23.5 | 37.09 | |
| Mini CPM2026.02 | 46 | 0 | 0 | 15 | 0 | 0 | 75 | 0 | 0 | 15.11 | |
| Mistral7BOCR=Tesseract2026.02 | 34 | 10 | 25 | 35 | 15 | 12.5 | 15 | 30 | 14.5 | 21.22 |