d-michail/Mitos-Qwen3-14B-Process

Mitos-Qwen3-14B-Process Overview

Mitos-Qwen3-14B-Process is a 14 billion parameter language model, fine-tuned from the Qwen/Qwen3-14B base model by d-michail. Its primary purpose is to provide accurate and structured answers to questions related to Greek e-government procedures, specifically drawing from the National Registry of Administrative Procedures (ΕΜΔΔ / Mitos). The model was trained on a comprehensive dataset of 90,145 question-answer records derived from all 4,407 procedures within this registry.

Key Capabilities

• Specialized Greek Public Administration QA: Optimized for question answering within the domain of Greek e-government, providing answers that reflect the structure and content of official registry entries.
• Structured and Detailed Responses: Generates longer answers that follow the format of the administrative registry, enhancing clarity and comprehensiveness.
• Research in QA: Intended for research in both closed-book and retrieval-augmented generation (RAG) settings for Greek public administration.
• High Context Length: Supports a maximum sequence length of 16384 tokens during training, allowing for processing of detailed queries and contexts.

Good For

• Researchers working on natural language processing for public administration, especially in the Greek context.
• Developing RAG systems for querying Greek e-government procedures, where accuracy and adherence to official documentation are crucial.
• Exploring domain-specific fine-tuning strategies for large language models on structured, factual datasets.

It's important to note that while the model learns the domain's register and format, its closed-book factual recall is limited; retrieval-augmented generation significantly improves accuracy. Answers are not legally authoritative and should be verified against live procedures.