A small, hand-curated rubric-graded evaluation set for measuring whether a language model can reason about the Nigerian economy the way an experienced Nigerian credit officer, SME owner, or independent analyst would.
This is v1 (seed), intentionally small. Each record is dense, with citations the model must use, omissions that lose points, a reference answer, and a per-record scoring rubric. The goal is to surface qualitative reasoning… See the full description on the dataset page:
https://huggingface.co/datasets/Apexgridapps/asotele-eval-nigerian-economy.