[KongchangAI]
Benchmark

AfriNORM

A scenario-based multiple-choice question dataset benchmark for measuring and diagnosing the normative reasoning ability and cultural competence of large language models in African sociocultural contexts.

Timeline (last 90 days)

Nov 18

AfriNORM is a benchmark for measuring and diagnosing the limitations of LLM normative reasoning in African sociocultural contexts, taking the form of a scenario-based multiple-choice dataset.

Unverified50%
Nov 18

The AfriNORM paper proposes two new metrics: the Universal Commonsense Bias Score (UCBS) and the Distractor Bias Score (DBS).

Unverified50%
Nov 18

AfriNORM currently uses only Yoruba as a case study, and the linguistic and cultural diversity of the African continent far exceeds what a single case can cover.

Unverified50%

All Facts (3)

Source Articles