VisCoR-55K is a high-quality visual reasoning dataset released in 2026 by Huazhong University of Science and Technology in collaboration with Alibaba Cloud. The dataset contains approximately 55,000 visual reasoning samples, each of which generates a corresponding reasoning process using comparative samples. It covers five major categories of high-quality visual reasoning datasets: general, reasoning, mathematical, graph, and OCR, and aims to promote research on reliable and robust visual reasoning using visual language models. Dataset composition – VQA Samples: Original Visual Question Answering Samples – Contrastive Counterparts: Matching question-and-answer pairs used to encourage credible reasoning. – Generated Rationales: High-quality inference chains synthesized using the VC-STaR framework

Citation

@inproceedings{pan2026through,
title={Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs},
author={Pan, Zhiyu and Wu, Yizheng and Hua, Jiasheng and Feng, Junyi and Yan, Shaotian and Deng, Bing and Cao, Zhiguo and Ye, Jieping},
booktitle={The Fourteenth International Conference on Learning Representations},
year={2026}
}