A developer has fine-tuned two open-source large language models, SILMA-9B-Darija and SILMA-2B-Darija, to better understand and generate Moroccan Arabic (Darija) and its informal Arabizi script. These models were trained on a custom dataset of over 46,000 Moroccan conversations and instructions, outperforming existing models like Gemma-2 and Llama-3.1 on benchmarks covering local vocabulary, history, and proverbs. AI
IMPACT Enhances LLM capabilities for underrepresented linguistic dialects, potentially improving accessibility and usability for specific user groups.
RANK_REASON Developer fine-tunes open-source models for a specific linguistic dialect, releasing them and the training data. [lever_c_demoted from research: ic=1 ai=1.0]
- abdnaouri/moroccan-darija-llm-dataset
- abdnaouri/silma-2b-darija-lora
- abdnaouri/silma-darija-9b-lora
- Gemma 2-2B
- GemMaroc-Qwen2.5-7B
- Llama-3.1:8b
- Moroccan Arabic
- Moroccan Darija Dataset
- MoroccanDarija-Llama-3.1-8B
- SILMA-2B-Darija
- SILMA-9B-Darija
- SILMA-AI/SILMA-9B-Instruct-v0.1
- SILMA-Darija
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →