Mushari440/qwen3-8B-SFT
Mushari440/qwen3-8B-SFT is an 8 billion parameter causal language model, fine-tuned from Qwen3-8B-Base by Mushari Alothman. This model is specifically optimized for supervised instruction following across both Arabic and English tasks, excelling in areas like MCQ answering, context-based QA, and general instruction adherence. It leverages bf16 mixed precision training on curated datasets to provide accurate and clean supervision.
Loading preview...
Mushari440/qwen3-8B-SFT: A Bilingual Instruction-Following Model
This model is an 8 billion parameter causal language model, developed by Mushari Alothman and fine-tuned from the Qwen3-8B-Base architecture. It is specifically designed for supervised instruction following in both Arabic and English, making it a versatile tool for bilingual applications.
Key Capabilities
- Bilingual Proficiency: Optimized for accurate supervision and instruction following in both Arabic and English.
- Question Answering: Excels at Multiple Choice Question (MCQ) answering in both languages.
- Contextual Understanding: Strong performance in context-based Question Answering (QA) and Retrieval-Augmented Generation (RAG) tasks.
- General Instruction Following: Capable of adhering to a wide range of general instructions.
Training Details
The model underwent Supervised Fine-Tuning (SFT) using bf16 mixed precision. The training data comprised curated Arabic and English datasets, including specific examples for MCQ, QA/RAG, and general instruction understanding. This focused training ensures its specialization in generating clean and accurate responses based on given instructions.
Good for
- Arabic and English MCQ answering.
- Context-based QA and RAG systems.
- General instruction following tasks where clear and accurate responses are needed.
Limitations
It is not intended for safety-critical or real-time decision-making, nor for generating factual guarantees without external verification.