Qolda_vision_benchmarks
Collection
These are the vision benchmarks used to evaluate Qolda model • 8 items • Updated
Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html>
<h"... is not valid JSON
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Dataset Summary
This is the human-translated Kazakh version of the AI2D (Allen Institute for AI Diagrams) dataset.
AI2D focuses on diagram parse and multi-choice question answering, specifically targeting elementary school science diagrams. This dataset has been human-translated, serving as a benchmark for evaluating how well multimodal models understand complex visual relationships and scientific concepts when prompted in Kazakh.
- Downloads last month
- 39