Visual Naturalness
Visual naturalness research focuses on understanding and replicating the qualities of naturally occurring images, particularly in the context of AI-generated content and adversarial attacks. Current efforts concentrate on developing methods to assess and improve the naturalness of AI-generated images and text, often employing generative models like GANs and VAEs, as well as analyzing attention mechanisms in transformer-based models. This research is crucial for advancing AI image and text generation, improving the robustness of AI systems against adversarial attacks, and ensuring the ethical and responsible use of AI in various applications.
Papers
Multi-perspective Alignment for Increasing Naturalness in Neural Machine Translation
Huiyuan Lai, Esther Ploeger, Rik van Noord, Antonio Toral
NAT-NL2GQL: A Novel Multi-Agent Framework for Translating Natural Language to Graph Query Language
Yuanyuan Liang, Tingyu Xie, Gan Peng, Zihao Huang, Yunshi Lan, Weining Qian