humain-ai/AraIFEval

---
language:
- ar
license: apache-2.0
tags:
- natural-language-understanding
pretty_naem: Arabic IFEval
configs:
- config_name: default
data_files: Arabic_IF_Eval.jsonl
- config_name: flat
data_files: Arabic_IF_Eval_Flat.jsonl
task_categories:
- text-generation
pretty_name: AraIFEval
size_categories:
- 1K<n<10K
---

## 🗂️ Dataset: AraIFEval
#### 📋 **Description:**
AraIFEval is an Arabic instruction-following (IF) evaluation benchmark developed to automatically measure language models’ adherence to specified instructions using verifiable techniques. The dataset includes 535 instances, each featuring two to four instructions that can be validated through deterministic programming. We compiled a set of 23 Arabic verifiable instructions, drawing inspiration from Zhou et al. (2023).
🌐 **Language(s)**: *Arabic* **|** 🧠 **Task Category**: *Text Generation*

<!-- | 🌐 Language(s)| 🧠 Task Category | Reference |
| ------------- | ------------------------- | ------------------------------- |
| Arabic | Multiple Choice Questions | **{{paper_title & Reference}}** | -->

#### 🔗 Hugging Face Usage Example
```python
from datasets import load_dataset
dataset = load_dataset("humain-ai/AraIFEval")
```

#### 📄 **Reference:**
[{{paper_title & Reference}}](https://www.example.com)