# Optical Character Recognition on internal multilingual OCR benchmarks

Source: [Optical Character Recognition on internal multilingual OCR benchmarks](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:47:55.745Z

Evaluation of Optical Character Recognition performance across multiple languages, including Korean, Japanese, French, German, Italian, Russian, Vietnamese, and Arabic, using an internal benchmark suite.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Optical Character Recognition |
| Dataset (including any stated split) | internal multilingual OCR benchmarks |
| Metric | Korean Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 2 of 2 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Korean Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Korean%20Accuracy>): Higher is better; unit: PERCENT.

- [Japanese Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Japanese%20Accuracy>): Higher is better; unit: PERCENT.

- [French Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=French%20Accuracy>): Higher is better; unit: PERCENT.

- [German Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=German%20Accuracy>): Higher is better; unit: PERCENT.

- [Italian Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Italian%20Accuracy>): Higher is better; unit: PERCENT.

- [Russian Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Russian%20Accuracy>): Higher is better; unit: PERCENT.

- [Vietnamese Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Vietnamese%20Accuracy>): Higher is better; unit: PERCENT.

- [Arabic Accuracy](<https://www.sota2.com/research/sota/optical-character-recognition-on-internal-multilingual-ocr-benchmarks.md?metric=Arabic%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Korean Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen2-VL-72B | 94.5 | Not recorded | 68fa75151e209131cea65976 |
| GPT-4o | 87.8 | Not recorded | 68fa75151e209131cea65976 |

### Qwen2-VL-72B

Korean Accuracy: 94.5. Result ID: 69938cbb3a6c31484916c67d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Implementation](<https://github.com/QwenLM/Qwen2-VL>)

Recorded evaluation configuration: {"Parameters":"72B"}. This record does not assert independent reproduction.

### GPT-4o

Korean Accuracy: 87.8. Result ID: 69938cbb3a6c31484916c67c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
