π About me
I am a Ph.D. candidate at the Hub of Computing and Data Science at the University of Hamburg, supervised by Prof. Dr. Chris Biemann.
My research focuses on large vision-language models and agentic AI systems, with particular interests in model training, multimodal learning and alignment, and systematic evaluation.
Publications
VIDA: A Dataset for Visually Dependent Ambiguity in Multimodal Machine Translation
VIDA introduces 2,500 visually grounded ambiguity instances and disambiguation-centric evaluation metrics for multimodal machine translation.
Chinese Toxic Language Mitigation via Sentiment Polarity Consistent Rewrites
ToxiRewriteCN studies sentiment-preserving detoxification with a new Chinese benchmark and an evaluation of 17 language models.
Rethinking Multilingual Vision-Language Translation: Dataset, Evaluation, and Adaptation
This work studies vision-language translation from the perspectives of data quality, model architecture, and evaluation, and proposes a balanced multilingual fine-tuning strategy.
CogSteer: Cognition-Inspired Selective Layer Intervention for Efficiently Steering Large Language Models
A cognition-inspired approach to selecting intervention layers for efficient, interpretable, and safer model steering.
Mitigating Hallucinations in Large Vision-Language Models with Instruction Contrastive Decoding
Instruction Contrastive Decoding mitigates object- and attribute-level hallucinations in large vision-language models during inference.
π₯ News
- May 2026: Released the VIDA dataset and preprint; the manuscript is under review at AACL 2026.
- Nov. 2025: Started my Ph.D. at the Hub of Computing and Data Science, University of Hamburg.
- Aug. 2025: ToxiRewriteCN was accepted to the main conference of EMNLP 2025.
- Jun. 2025: Released an arXiv preprint on multilingual vision-language translation.
- May 2025: CogSteer was accepted to Findings of ACL 2025.
- May 2024: Instruction Contrastive Decoding was accepted to Findings of ACL 2024.
ποΈ Professional Activities
Reviewing Service
- Conference reviewer for ACL, NAACL, ACM Multimedia, EMNLP, LREC, CVPR, and KONVENS (2024βpresent).
Teaching
- Lecturer, Exercises for Statistical Methods in Language Technology, University of Hamburg (Summer 2025 and 2026).
- Lecturer, Exercises for Natural Language Processing and the Web, University of Hamburg (Winter 2025/26).
πΊ Beyond research
I enjoy craft beer and exploring the stories behind local breweries.