publications

2025

1. EMNLP

   

   Multimodal Language Models See Better When They Look Shallower

   Haoran Chen, Junyan Lin, Xinhao Chen, Yue Fan, Xin Jin, Hui Su, Jianfeng Dong, Jinlan Fu, and Xiaoyu Shen

   In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025

   Main Conference, Oral Presentation

   Abs Bib HTML Video Poster

   A systematic study of how visual layer selection affects multimodal language model performance.

   @inproceedings{chen2025shallower,
     title = {Multimodal Language Models See Better When They Look Shallower},
     author = {Chen, Haoran and Lin, Junyan and Chen, Xinhao and Fan, Yue and Jin, Xin and Su, Hui and Dong, Jianfeng and Fu, Jinlan and Shen, Xiaoyu},
     booktitle = {Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing},
     year = {2025},
     note = {Main Conference, Oral Presentation},
   }

2. CVPR

   

   Multi-Layer Visual Feature Fusion in Multimodal LLMs: Methods, Analysis, and Best Practices

   Junyan Lin^*, Haoran Chen^*, Yue Fan, Yingqi Fan, Xin Jin, Hui Su, Jinlan Fu, and Xiaoyu Shen

   In Proceedings of the Computer Vision and Pattern Recognition Conference, 2025

   Abs Bib HTML PDF

   An analysis of visual layer selection and fusion strategies for multimodal large language models.

   @inproceedings{lin2025multilayer,
     title = {Multi-Layer Visual Feature Fusion in Multimodal LLMs: Methods, Analysis, and Best Practices},
     author = {Lin, Junyan and Chen, Haoran and Fan, Yue and Fan, Yingqi and Jin, Xin and Su, Hui and Fu, Jinlan and Shen, Xiaoyu},
     booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference},
     pages = {4156--4166},
     year = {2025},
   }

2024

1. EMNLP

   

   To Preserve or To Compress: An In-Depth Study of Connector Selection in Multimodal Large Language Models

   Junyan Lin, Haoran Chen, Dawei Zhu, and Xiaoyu Shen

   In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024

   Abs arXiv Bib HTML

   A comprehensive study of connector selection strategies in multimodal large language models.

   @inproceedings{lin2024preserve,
     title = {To Preserve or To Compress: An In-Depth Study of Connector Selection in Multimodal Large Language Models},
     author = {Lin, Junyan and Chen, Haoran and Zhu, Dawei and Shen, Xiaoyu},
     booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing},
     year = {2024},
   }