---
pretty_name: albert_OCR_v1
configs:
  - config_name: albert_ocr_v2__1-1121
    data_files: data/albert_ocr_v2__1-1121/*.parquet
    default: true
  - config_name: albert_ocr_v2__0-1120
    data_files: data/albert_ocr_v2__0-1120/*.parquet
tags:
  - evalap
  - evaluation
  - llm
---

# albert_OCR_v1 (ID: 60)

Evaluating OCR capabilities of albert on the marker datasets

## Overview

This dataset contains **2 experiments** 
from the EvalAP evaluation platform.

**Datasets:** OCR_marker_benchmark

**Models evaluated:** mistralai/Mistral-Small-3.1-24B-Instruct-2503

**Metrics:** generation_time, ocr_v1, output_length


## Scores

### OCR_marker_benchmark

| model                                         | generation_time   | ocr_v1        | output_length   |
|:----------------------------------------------|:------------------|:--------------|:----------------|
| mistralai/Mistral-Small-3.1-24B-Instruct-2503 | 35.95 ± 38.47     | 81.59 ± 21.33 | 424.21 ± 248.22 |



## Usage

Use the dropdown above to select an experiment configuration.