zox-BT/gemma-2b-cameroon-cultural-blindspots

Gemma-2b Cameroon Cultural Blindspots

This dataset highlights the "blind spots" of the Google Gemma-2-2b base model regarding Cameroonian culture, geography, and local languages.

1. Model Tested

• —Model Name: google/gemma-2-2b
• —Type: Base Model (Pre-trained)

2. Loading Procedure

The model was loaded using the transformers library on a Google Colab T4 GPU:

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "google/gemma-2-2b"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

3. Analysis & Proposed Fix

The model consistently fails on local cultural nuances (traditional dishes, specific geography, and ethnic names) and native languages like Medumba.

How to fix this:

• —Dataset type: I recommend a SFT (Supervised Fine-Tuning) dataset with a focus on instruction-following for African cultural contexts.
• —Assembly: I would crawl local academic repositories (ENSPY papers), digitize cultural archives from the West Region, and use "Human-in-the-loop" verification by native speakers.
• —Dataset Size: A curated set of 5,000 to 10,000 high-quality samples is enough to significantly reduce these hallucinations without degrading the model's general reasoning.