qu-llm-assistant-allam

QU-LLM Assistant (ALLaM) — wesam3/qu-llm-assistant-allam

A fine-tuned ALLaM-7B-Instruct model specialised in answering student advisory questions for Qassim University (جامعة القصيم). ALLaM is an Arabic-first large language model developed by SDAIA (Saudi Data and Artificial Intelligence Authority) and is particularly strong on Arabic-language tasks.

The model is distributed as a Q4_K_M GGUF file for efficient local inference with llama.cpp / Ollama.

---

Model Details

---

Training Data

• Dataset: 12,320 curated Q&A pairs filtered exclusively from 2024–2026 Qassim University documents — the most current bylaws, handbooks, and policy circulars.
• Source: Publicly available PDFs and web pages from qu.edu.sa.
• Language: Predominantly Arabic.
• Rationale for filtering: Focusing on 2024-2026 documents ensures the model reflects the latest university regulations and avoids outdated policies.

---

Usage

Ollama (recommended)

ollama run wesamhamad/qu-llm-assistant-allam

Or pull the GGUF manually and create a local Modelfile:

FROM ./ALLaM-7B-Q4_K_M.gguf
SYSTEM "أنت مساعد ذكاء اصطناعي متخصص في الإجابة على استفسارات طلاب جامعة القصيم."

ollama create qu-llm-assistant-allam -f Modelfile
ollama run qu-llm-assistant-allam

llama.cpp

llama-cli -m ALLaM-7B-Q4_K_M.gguf \
  -p "ما هي متطلبات التسجيل في الفصل الدراسي الأول؟"

Python (llama-cpp-python)

from llama_cpp import Llama

llm = Llama(model_path="ALLaM-7B-Q4_K_M.gguf")
response = llm.create_chat_completion(messages=[
    {"role": "system",  "content": "أنت مساعد أكاديمي لجامعة القصيم."},
    {"role": "user",    "content": "ما هي شروط الحصول على المنحة الدراسية؟"},
])
print(response["choices"][0]["message"]["content"])

---

Why ALLaM?

ALLaM was pre-trained on an extensive Arabic corpus by SDAIA, giving it superior morphological understanding and dialectal coverage compared with multilingual models of similar size. Fine-tuning ALLaM on Qassim University data yields a model that is both culturally aware and domain-specific.

---

Intended Use & Limitations

• Intended use: Helping students navigate university regulations, registration procedures, scholarship requirements, and academic policies at Qassim University — using the most up-to-date (2024-2026) policy documents.
• Out-of-scope: General-purpose chat, medical/legal advice, topics unrelated to the university.
• Hallucinations: Like all LLMs the model may occasionally produce incorrect information. Always verify critical details with the official university portal.

---

Citation

@misc{qu-llm-assistant-allam-2025,
  author    = {Wesam Hamad},
  title     = {QU-LLM Assistant (ALLaM): Fine-tuned ALLaM-7B for Qassim University Advisory},
  year      = {2025},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/wesam3/qu-llm-assistant-allam}
}

---

License

Apache-2.0 — see LICENSE.