OpenworldAUC: Towards Unified Evaluation and Optimization for Open-world Prompt Tuning

Cong Hua, Qianqian Xu^*, Zhiyong Yang, Zitai Wang, Shilong Bao, Qingming Huang^*

International Conference on Machine Learning (ICML) 2025

[Paper] |[Code] | [Project]

• This paper proposes OpenworldAUC, a metric for Open-World Prompt Tuning, unifying base-to-new detection, base/new classification, and robustness to domain shifts. The Gated Mixture-of-Prompts framework jointly optimizes OpenworldAUC via multi-prompt collaboration under a divide-and-conquer strategy, supported by theoretical guarantees.

ReconBoost: Boosting Can Achieve Modality Reconcilement

Cong Hua, Qianqian Xu^*, Shilong Bao, Zhiyong Yang, Qingming Huang^*

International Conference on Machine Learning (ICML) 2024

[Paper] |[Code] | [Project]

• This paper explores a novel multi-modal alternating learning paradigm pursuing a reconciliation between the exploitation of uni-modal features and the exploration of cross-modal interactions.

Hierarchical Set-to-set Representation for 3D Cross-modal Retrieval

Yu Jiang, Cong Hua, Yifan Feng, Yue Gao^*. (Student first author)

IEEE Transactions on Neural Networks and Learning Systems (IEEE TNNLS) 2023

[Paper] |[Code] | [Project]

• This paper proposes a hierarchical set-to-set representation and a corresponding hierarchical similarity that incorporates global-to-global and local-to-local similarity metrics for 3-D object cross-modal retrieval.

Efficient Adversarial Training for AUC Optimization

Shilong Bao, Qianqian Xu^*, Zhiyong Yang, Cong Hua, Boyu Han, Xiaochun Cao, Qingming Huang^*. Chinese Journal of Computers 2025

[Paper]

• This paper presents a systematic investigation of a more generalized and efficient AdAUC framework, introducing a novel Ranking-aware Adversarial Regularization algorithm to achieve a better balance between standard and robust AUC performance.

SHREC'22 Track: Open-Set 3D Object Retrieval

$28$ authors including Cong Hua. Computers and Graphics 2022

[Paper] |[Track]

• This paper reports the results of the SHREC’22 track: Open-Set 3D Object Retrieval, the goal of which is to evaluate the performance of different retrieval algorithms under the Open-Set setting and modality-missing setting, respectively.