Dongzhi Jiang

I'm a PhD student in Multimedia Lab, CUHK, supervised by Prof. Hongsheng Li.

Email  /  Google Scholar  /  Github

profile photo

Selected Publication

I am interested in AIGC. Currently, I am focusing on Multimodal Large Language Model (MLLM) and Text-to-Image Diffusion models.

Logo MMSearch: Benchmarking the Potential of Large Models as Multi-modal Search Engines
Dongzhi Jiang*, Renrui Zhang*, Ziyu Guo, Yanmin Wu, Jiayi Lei, Pengshuo Qiu, Pan Lu, Zehui Chen, Guanglu Song, Peng Gao, Yu Liu, Chunyuan Li, Hongsheng Li,
arxiv, 2024
arXiv / website / dataset / GitHub

We investigate the potential of current LMM to function as a multimodal AI search engine. We also introduce a multimodal AI search engine pipeline and outperforms Perplexity-pro with only open-source LMMs.

💫CoMat: Aligning Text-to-Image Diffusion Model with Image-to-Text Concept Matching
Dongzhi Jiang, Guanglu Song, Xiaoshi Wu, Renrui Zhang, Dazhong Shen, Zhuofan Zong, Yu Liu, Hongsheng Li
Neurips, 2024
arXiv / website / GitHub

A fine-tuning strategy to address the text-to-image misalignment issue with image-to-text concept matching. The training data only includes text prompts.

Logo MathVerse: Does Your Multi-modal LLM Truly See the Diagrams in Visual Math Problems?
Renrui Zhang*, Dongzhi Jiang*, Yichi Zhang*, Haokun Lin, Ziyu Guo, Pengshuo Qiu, Aojun Zhou, Pan Lu, Aojun Zhou, Kai-Wei Chang, Peng Gao, Hongsheng Li
ECCV, 2024
arXiv / website / dataset / GitHub

We investigate current benchmarks to incorporate excessive visual content within textual questions, which potentially assist MLLMs in deducing answers without truly interpreting the input diagrams.

Temporal Enhanced Training of Multi-view 3D Object Detector via Historical Object Prediction
Zhuofan Zong*, Dongzhi Jiang*, Guanglu Song, Zeyue Xue, Jingyong Su, Hongsheng Li, Yu Liu
ICCV, 2023
arXiv / GitHub

We design a plug-and-play approach to enhance the temporal modeling capability of BEV detectors with no additional inference cost.


The source code of this website is adapted from Jon Barron's website.