Qolda_vision_benchmarks
Collection
These are the vision benchmarks used to evaluate Qolda model • 8 items • Updated
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.
MathVision-Kazakh is a machine-translated dataset of the MathVision dataset into the Kazakh language. This dataset is designed to evaluate the mathematical reasoning and visual understanding capabilities of Multimodal Large Language Models within a specific linguistic context.
| Field | Description |
|---|---|
index |
Unique identifier corresponding to the original MathVision entry. |
question |
The mathematical problem statement translated into Kazakh. |
options |
A list of choices for the question (translated where applicable, formulas preserved). |
image |
Path to the associated image file. |
answer |
The correct choice (A, B, C, D,). |
level |
Difficulty level ranging from 1 to 5. |
Below is an example of a typical data entry in the dataset:
{
"index": 105,
"question": "Берілген суреттегі шеңбердің радиусы 5 см болса, боялған бөліктің ауданын табыңыз (π = 3.14).",
"options": "[\"25.5\", \"31.4\", \"39.25\", \"78.5\"]",
"image": "images/105.jpg",
"answer": "C",
"level": "3"
}