β¨ The paper Local Manifold Sensitivity and Regularization in Long-Tailed Recognition was accepted to ACM MM 2026 Main Track.
π Recognized as an ICML 2026 Golden Reviewer.
π Two papers led by master's students I helped mentor were accepted to IJCNN 2026 Main Track.
β¨ The paper CurrMix: Curriculum-Enhanced MixUp for Long-Tailed Visual Recognition was accepted to CVPR 2026 Findings.
β¨ The paper Prototype Entropy Alignment: Reinforcing Structured Uncertainty in LLM Reasoning was accepted to AAAI 2026 Main Track as an Oral presentation.
β¨ The paper Supervised Exploratory Learning for Long-Tailed Visual Recognition was accepted to ICCV 2025 Main Track.
β¨ The paper Enhancing Mixture of Experts with Independent and Collaborative Learning for Long-Tail Visual Recognition was accepted to IJCAI 2025 Main Track.
Research Interests
My work focuses on representation learning under imbalance, cross-modal alignment, and trustworthy reasoning for reliable AI systems.
Long-tailed Visual RecognitionMultimodal LearningTrustworthy AI Reasoning
Long-tailed Visual RecognitionRepresentation learning under class imbalance, with emphasis on tail-class robustness and decision calibration.
Multimodal LearningVision-language alignment, retrieval-augmented modeling, and multimodal sentiment or reasoning pipelines.
Trustworthy AI ReasoningUncertainty-aware reasoning, structured prediction, and evaluation mechanisms for reliable language and multimodal AI systems.
Education
2017.09-2021.06Nanjing University of Posts and Telecommunications, B.Sc. in Opto-electronic Information Science and Engineering
2021.09-2022.11University of Southampton, M.Sc. in Electronic Engineering
2023.09-PresentXiamen University, PhD in Computer Science
Publications
Work in long-tailed visual recognition, multimodal learning, and language-model reasoning.
2026
Local Manifold Sensitivity and Regularization in Long-Tailed Recognition
Y. Chen*, Z. Pan*, S. Hu, J. Jiao, X. Chen, Z. Jian, Q. Wu.
ACM MM 2026 Main Track (CCF-A)
Paper coming soon
2025
Enhancing Mixture of Experts with Independent and Collaborative Learning for Long-Tail Visual Recognition
Y. Chen*, Z. Jian*, N. Ke, S. Hu, J. Jiao, Q. Hong, Q. Wu.
AEM45K: A Large-scale Multimodal Dataset and Method for Aesthetic Assessment of Student Artwork in Middle-school Education
S. Hu*, Y. Chen*, Y. Liu, R. Liang, J. Jiao, J. Jin, W. Liu, Y. Zhang, L. Wang, Q. Wu.
IJCNN 2026 Main Track (CCF-C)
Paper coming soon
2026
Tackling Mask Imbalance: Prototypical Mixture-of-Experts with Hardness-Aware Mining for Medical Segmentation
J. Jiao, G. Liu, Y. Chen, S. Hu, W. Liu, Y. Zhang, L. Wang, Q. Wu.
IJCNN 2026 Main Track (CCF-C)
Paper coming soon
* denotes equal contribution.
Selected Projects
Applied AI projects connecting intelligent assessment, aesthetic education, and data-driven educational evaluation.
Shaping Individual Aesthetic Life: Intelligent Assessment and Application of Art Creativity-Emotion Competence
Algorithm Model Lead
Led the design and implementation of intelligent assessment models for students' aesthetic development, covering creativity-emotion indicators, artwork analysis, classroom interaction, and teacher feedback. The platform was selected as a national intelligent aesthetic-education application case.
Digital Intelligence for K-12 Aesthetic Education Evaluation: A Foshan Case Study
Recommendation Algorithm Developer
Developed recommendation algorithms for a regional aesthetic-education evaluation platform, supporting process-based, multidimensional monitoring and data-driven diagnosis of resource allocation, teacher support, and teaching quality.
Academic Service
π Reviewer for CCF-A conferences, including ICML, NeurIPS, and ACM MM.
.JPG)
Long-tailed Visual Recognition Representation learning under class imbalance, with emphasis on tail-class robustness and decision calibration.
Multimodal Learning Vision-language alignment, retrieval-augmented modeling, and multimodal sentiment or reasoning pipelines.
Trustworthy AI Reasoning Uncertainty-aware reasoning, structured prediction, and evaluation mechanisms for reliable language and multimodal AI systems.
