Greek e-government Q&A: the mitos-qa dataset and Qwen3 models (1.7B/8B/14B; Dialogue and Process variants; merged and LoRA) fine-tuned on it.