1.chinese_simple_vqa.jsonl（the image is in url format）
2.chinese_simplevqa.parquet (the image is in base64 format and can be downloaded)
Chinese SimpleVQA is the first factuality-based visual question-answering benchmark in Chinese, aimed at assessing the visual factuality of LVLMs across 8 major topics and 56 subtopics. The key features of this benchmark include a focus… See the full description on the dataset page:
https://huggingface.co/datasets/OpenStellarTeam/Chinese-SimpleVQA.
