Evidence
Check the work yourself
Our claims are tied to a specific benchmark and a published method. Where a result is benchmark-specific, we say so.
Preprint
arXiv:2606.20255
The register framework, benchmark construction and full results. A preprint — not peer-reviewed.
Press
TechEconomy coverage
Independent reporting on the register gap in frontier AI systems.
Credentials
Verify a CMIA credential
Check any credential issued through Cihan Digital Academy against its signature.
Nigerian register benchmark
| Gemini Flash | 78.6 |
| GPT-5 | 67.8 |
| Gemini Pro | 65.4 |
| Calibrated human analysts | 100.0 |
MIS™ scores on our Nigerian register/discourse benchmark using a contamination-aware holdout. Human performance demonstrated on this benchmark only; these figures are not a general model ranking.