3MegaLabs Research

Research

Peer-reviewable publications, audit results, and benchmarking roadmap. Every claim is testable. Every result is reproducible.

Independent Audit Results

Gemini 2.5 Flash independently reviewed the Minimum Representative Sample (MRS) for Bemba tonal data.

132 Records — 100% Tonal Accuracy
Meeussen's Rule 48/48
Binary Spreading 82/82
Nasal Harmony 22/22
Melodic Override 26/26

Benchmarking Roadmap

Planned evaluations against established Bantu language benchmarks.

PLANNED
BembaSpeech
ASR benchmark — evaluate whether BantuNomics tonal data improves speech recognition accuracy for Bemba.
PLANNED
AfriQA
Question answering — test whether morphological decomposition improves comprehension in Bantu languages.
PLANNED
MasakhaPOS
Part-of-speech tagging — evaluate morpheme-level data as pre-training for POS taggers across multiple Bantu languages.

Citation

@techreport{cintu2026bantunomics,
  title     = {BantuNomics: The Tonal Operating System for Bantu Language AI},
  author    = {Cintu, Conti and others},
  year      = {2026},
  institution = {3MegaLabs},
  url       = {https://bantunomics.com/research}
}