Pinned Loading

1. Amharic-LoRA-Experiment Amharic-LoRA-Experiment Public

   Small LoRA experiment adapting Qwen2.5-0.5B to Amharic, evaluated with bits per character on FLORES-200

   Jupyter Notebook

2. Amharic-Tokenization-Benchmark Amharic-Tokenization-Benchmark Public

   Benchmark of tokenizer efficiency (fertility, chars per token) for Amharic vs English on FLORES-200 across 9 LLM tokenizers

   Jupyter Notebook

3. Amharic-LLM-RAG-Benchmark Amharic-LLM-RAG-Benchmark Public

   Empirical Evaluation of Retrieval Paradigms and Model Specialization for Low-Resource Amharic Question Answering

   Jupyter Notebook

4. anc-rag-assistant anc-rag-assistant Public

   Jupyter Notebook

5. Simple-Amharic-RAG-System Simple-Amharic-RAG-System Public

   A small Retrieval-Augmented Generation (RAG) pipeline for Amharic, comparing embedding models and evaluated with RAGAS

   Jupyter Notebook

6. Amharic-Subword-Tokenization-Analysis Amharic-Subword-Tokenization-Analysis Public

   Empirical analysis of subword tokenization algorithms (BPE, WordPiece, Unigram) for Amharic and Ge'ez script. Benchmarks cross-lingual "Token Tax" against foundation LLMs (LLaMA-3, Mistral, GPT-4o)…

   Jupyter Notebook