RISEN: Incremental Multilingual Text Recognition by Sharing and Fusing Cross-language Knowledge
Incremental multilingual text recognition through sharing and fusing cross-language knowledge.
About Xiao-Qian Liu
I am a postdoctoral fellow at the School of Software, Shandong University. I received my Ph.D. in Artificial Intelligence from Shandong University in 2024, graduating six months early, and my M.S. in Control Engineering in 2020.
Research
My research focuses on Trustworthy AI, LVLM Reasoning, Multimedia Analysis, and Document Intelligence. My long-term goal is to develop generalizable, reliable, and controllable frameworks for trustworthy reasoning in LVLMs by investigating the fundamental principles of model optimization and representation learning.
My current work investigates how visual evidence, textual prompts, and generated history jointly contribute to token prediction.
Research Interests
Trustworthy AI, Multimodal Large Language Models, Document Intelligence, Continual Cross-Lingual Learning, Scene Text Recognition, and Reliable Learning under Distribution Shift.
“My long-term research goal is to develop generalizable, reliable, and controllable frameworks for trustworthy reasoning in LVLMs.”
Selected Publications
View all → Self-Supervised Discovery of Cross-Lingual Shared Knowledge for Continual Text Recognition
Continual multilingual text recognition through structured discovery and transfer of shared cross-lingual knowledge.
Hierarchical Multi-label Learning for Incremental Multilingual Text Recognition
Hierarchical multi-label learning for incremental multilingual text recognition.
Noisy-Aware Unsupervised Domain Adaptation for Scene Text Recognition
Unsupervised adaptation for scene text recognition with learning mechanisms that suppress unreliable pseudo-label supervision.
ProtoUDA: Prototype-Based Unsupervised Adaptation for Cross-Domain Text Recognition
Prototype-based unsupervised adaptation for robust text recognition across domains.
TextAdapter: Self-Supervised Domain Adaptation for Cross-Domain Text Recognition
Self-supervised representation adaptation for cross-domain scene text recognition.
Unsupervised Domain Adaptation via Class Aggregation for Text Recognition
Class-level aggregation for unsupervised domain adaptation in text recognition.
Meta-Learning Based Prototype-Relation Network for Few-Shot Classification
A prototype-relation network based on meta-learning for few-shot classification.
Research Projects
CAST Young Talent Support Program (Doctoral Special Program)
Principal Investigator · 2024
China Postdoctoral Science Foundation
Principal Investigator · 2026
Postdoctoral Fellowship Program (Grade B) of China Postdoctoral Science Foundation
Principal Investigator · 2025
Shandong Province Postdoctoral Innovation Project (First-Class)
Principal Investigator · 2026