Capturing data…
Capturing data…
FRI, 6 NOV · 97 ITEMS
"NaijaFacts benchmark separates language from knowledge about Nigeria" — arXiv cs.CL · AI & Technology
is a benchmark framework introduced in a recent preprint that aims to evaluate how well language models separate language comprehension from factual knowledge specifically about Nigeria. The researchers, led by Zainab Afolabi, test models on questions asked in both English and several Nigerian languages to distinguish whether poorer performance is due to language barriers or gaps in knowledge about Nigerian facts.
The claim is drawn directly from the paper itself, posted on arXiv () as a primary source. The benchmark design and initial results are described in the preprint, so the accuracy is as solid as the paper's pre-publication claims.
The paper was published on arXiv on November 5, 2026, making this genuinely new material.