Browse Datasets

IrokoBench

Comprehensive Benchmark for African Languages

benchmarking evaluation multi-task Multiple evaluation sets CC-BY-4.0
Access Dataset

Description

A comprehensive benchmark suite for evaluating LLMs on African languages across multiple tasks. Covers translation, classification, QA, and more for systematic model evaluation.

Quick Start

Python 
from datasets import load_dataset

# Load specific task subset
dataset = load_dataset("masakhane/irokobench", "yor_classification", split="test")

Project Ideas

LLM Evaluation

Benchmark popular LLMs (GPT-4, Llama, Gemini) on African languages

Intermediate

Quantization Impact

Measure accuracy degradation when quantizing models using IrokoBench

Advanced

Ethical Considerations

  • Standardized evaluation helps identify model biases