IrokoBench: A New Benchmark for African Languages in the Age of Large Language Models

Despite the widespread adoption of Large language models (LLMs), their remarkable capabilities remain limited to a few high-resource languages. Additionally, many low-resource languages (e.g. African languages) are often evaluated only on basic text classification tasks due to the lack of appropriat...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Adelani, David Ifeoluwa, Ojo, Jessica, Israel Abebe Azime, Jian Yun Zhuang, Alabi, Jesujoba O, He, Xuanli, Ochieng, Millicent, Hooker, Sara, Bukula, Andiswa, En-Shiun Annie Lee, Chukwuneke, Chiamaka, Buzaaba, Happy, Sibanda, Blessing, Godson Kalipe, Mukiibi, Jonathan, Kabongo, Salomon, Yuehgoh, Foutse, Setaka, Mmasibidi, Ndolela, Lolwethu, Odu, Nkiruka, Mabuya, Rooweither, Shamsuddeen Hassan Muhammad, Osei, Salomey, Samb, Sokhar, Tadesse Kebede Guge, Stenetorp, Pontus
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 05.06.2024
Subjects
African languages
Benchmarks
English language
Large language models
Natural language processing
Proprietary
Test sets
Translating
Online AccessGet full text

Cover

More Information
Summary:Despite the widespread adoption of Large language models (LLMs), their remarkable capabilities remain limited to a few high-resource languages. Additionally, many low-resource languages (e.g. African languages) are often evaluated only on basic text classification tasks due to the lack of appropriate or comprehensive benchmarks outside of high-resource languages. In this paper, we introduce IrokoBench -- a human-translated benchmark dataset for 16 typologically-diverse low-resource African languages covering three tasks: natural language inference~(AfriXNLI), mathematical reasoning~(AfriMGSM), and multi-choice knowledge-based QA~(AfriMMLU). We use IrokoBench to evaluate zero-shot, few-shot, and translate-test settings~(where test sets are translated into English) across 10 open and four proprietary LLMs. Our evaluation reveals a significant performance gap between high-resource languages~(such as English and French) and low-resource African languages. We observe a significant performance gap between open and proprietary models, with the highest performing open model, Aya-101 only at 58\% of the best-performing proprietary model GPT-4o performance. Machine translating the test set to English before evaluation helped to close the gap for larger models that are English-centric, like LLaMa 3 70B. These findings suggest that more efforts are needed to develop and adapt LLMs for African languages.
ISSN:2331-8422