A retinal generative foundation model that generates dynamic fluorescein angiography videos from single color fundus images, with demonstrated downstream transferability across multiple external ophthalmology datasets.
@article{zhang2026fundus,title={Fundus to Fluorescein Angiography Video Generation as a Retinal Generative Foundation Model},author={Zhang, Weiyi and Yang, Jian and Chen, Ruoyu and Huang, Siyu and Xu, Pusheng and Chen, Xiaolan and Wang, Ziwei and Lu, Shanfu and Cao, Hongyu and others},journal={npj Digital Medicine},year={2026},}
Preprint
Parking Without Building: Agentic AI for Parking Planning Through Urban Contradiction Reasoning
Hongyu Cao, David A. King, Xinyuan Wang, and 4 more authors
An agentic AI framework for parking planning that reasons over urban contradictions to allocate parking capacity without new construction.
@article{cao2026parking,title={Parking Without Building: Agentic AI for Parking Planning Through Urban Contradiction Reasoning},author={Cao, Hongyu and King, David A. and Wang, Xinyuan and Malarkkan, Arun Vignesh and Liu, Kunpeng and Wang, Dongjie and Fu, Yanjie},journal={Preprint},year={2026},}
arXiv
Mitigating Shortcut Reasoning in Language Models: A Gradient-Aware Training Approach
Hongyu Cao, Kunpeng Liu, Dongjie Wang, and 1 more author
Shortcut-Aware Reasoning Training (SART), a gradient-aware framework that detects and mitigates shortcut-promoting samples via a shortcut score and gradient surgery, improving reasoning generalization under distribution shifts.
@article{cao2026mitigating,title={Mitigating Shortcut Reasoning in Language Models: A Gradient-Aware Training Approach},author={Cao, Hongyu and Liu, Kunpeng and Wang, Dongjie and Fu, Yanjie},journal={arXiv preprint arXiv:2603.20899},year={2026},}
arXiv
Sim2Act: Robust Simulation-to-Decision Learning via Adversarial Calibration and Group-Relative Perturbation
Hongyu Cao, Jinghan Zhang, Kunpeng Liu, and 5 more authors
A robust simulation-to-decision framework that re-weights simulation errors in decision-critical state-action pairs through adversarial calibration and stabilizes policy preferences with a group-relative perturbation strategy.
@article{cao2026sim2act,title={Sim2Act: Robust Simulation-to-Decision Learning via Adversarial Calibration and Group-Relative Perturbation},author={Cao, Hongyu and Zhang, Jinghan and Liu, Kunpeng and Wang, Dongjie and Xia, Feng and Chen, Haifeng and Hu, Xia and Fu, Yanjie},journal={arXiv preprint arXiv:2603.09053},year={2026},}
2025
BigData
Structured Memory and Role-Aware Decision Making for Supply Chain Transportation
Hongyu Cao, Haoyue Bai, and Yanjie Fu
In 2025 IEEE International Conference on Big Data (BigData), 2025
A decision-making method for supply chain transportation that combines structured memory with role-aware reasoning to improve large language model decisions.
@inproceedings{cao2025structured,title={Structured Memory and Role-Aware Decision Making for Supply Chain Transportation},author={Cao, Hongyu and Bai, Haoyue and Fu, Yanjie},booktitle={2025 IEEE International Conference on Big Data (BigData)},pages={1837--1846},year={2025},}
Information
Desentiment: A New Method to Control Sentimental Tendency During Summary Generation
Desentiment, a method that controls the sentimental tendency during text summarization, enabling generation of summaries with desired sentiment properties.
@article{cao2025desentiment,title={Desentiment: A New Method to Control Sentimental Tendency During Summary Generation},author={Cao, Hongyu and Li, Jinlong},journal={Information},volume={16},number={6},pages={453},year={2025},}