SIB-200

Benchmark for 200+ Languages

topic-classification benchmarking evaluation Based on FLORES-200 CC-BY-SA 4.0

Description

Large-scale benchmark for topic classification in 200+ languages. Excellent for benchmarking compressed models.

Quick Start

Python
from datasets import load_dataset

dataset = load_dataset("Davlan/sib200", "hau_Latn", split="test")

Project Ideas

Quantization Benchmark

Quantize AfroXLMR to 8-bit and benchmark accuracy loss

Intermediate

Speed Comparison

Compare mBERT vs DistilmBERT inference speed

Beginner

Ethical Considerations

  • None significant