Yana/ft-llm-2026-qa-dataset

FT-LLM 2026 QA Dataset A Japanese visual-question-answering dataset used for Stage 1-2 visual instruction tuning of the COMPASS Vision-Language Model. Each sample contains a document or natural image together with one or more Japanese question–answer pairs, and is designed to give the VLM its instruction-following and VQA capabilities. Images are embedded in the dataset, so no external downloads are required. Part of the Compass collection. License Released under… See the full description on the dataset page: https://huggingface.co/datasets/Yana/ft-llm-2026-qa-dataset.