haikal1623/qwen2.5-7b-legal-id-sft
The haikal1623/qwen2.5-7b-legal-id-sft model is a 7.6 billion parameter Qwen2.5-7B-Instruct variant, fine-tuned by Haikal Fairuzi Maulana specifically for answering questions about Indonesian labor law in Bahasa Indonesia. This supervised fine-tuned model is the first stage in a pipeline designed to run on consumer hardware, focusing on specialized legal domain knowledge. It excels at providing information on Indonesian labor regulations, serving as a foundational component for legal information retrieval systems.
Loading preview...
haikal1623/qwen2.5-7b-legal-id-sft: Indonesian Labor Law LLM
This model is a 7.6 billion parameter Qwen2.5-7B-Instruct variant, developed by Haikal Fairuzi Maulana, specifically fine-tuned for Indonesian labor law (hukum ketenagakerjaan) questions. It represents the first stage of a three-stage pipeline, with subsequent stages involving reasoning reinforcement learning and retrieval-augmented generation (RAG).
Key Capabilities & Features
- Specialized Domain: Expert in Indonesian labor law, answering questions in Bahasa Indonesia.
- Consumer Hardware Optimized: Designed to run efficiently on consumer-grade GPUs, such as an 8 GB RTX 4060 Ti.
- Training Method: Utilizes LoRA via Unsloth and TRL on an Indonesian instruction dataset in ChatML format.
- Pipeline Component: Intended to be used as part of a larger system, typically with a retrieval layer to verify answers against actual regulations.
Intended Use & Limitations
This model is designed for answering questions about Indonesian labor-law regulations. It is crucial to note that it is not a lawyer and its output is not legal advice. Users should treat every answer as a draft requiring verification against source regulations. The model's knowledge is bounded to its training data, lacking awareness of regulations outside that set or amendments made after training. Performance in languages other than Bahasa Indonesia is untested, and no formal legal-accuracy benchmark has been conducted.