Jinxu Zhang

Jinxu Zhang

Auto-generated profile, based on publicly available information

Jinxu Zhang

Harbin Institute of Technology, Harbin, China

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 November 2023 in IEEE Transactions on Learning Technologies
Verified
Journal Article
Closed access
Ming liuMing liu
Ming liu
JZJingxu Zhang
Jingxu Zhang
LNLucy Michael Nyagoga
Lucy Michael Nyagoga
LLLi Liu
Li Liu
Student question generation (SQG) is an effective strategy for improving reading comprehension. It helps students improve their understanding of reading materials, metacognitively monitor their comprehension, and self-correct comprehension gaps. Internet technologies have been used to facilitate student question generation process through intensive peer support. However, the availability, level of task commitment, and capabilities of student peers have emerged as significant concerns, particularly in light of the global pandemic and the subsequent post-pandemic era. Thus, this article presents a student-Artificial Intelligence (AI) co-creation tool called CoAsker for supporting question generation. Following recent Human Computer Interaction (HCI) research in human-AI collaborative writing, CoAsker first allows students to provide question clues and answers and then uses a state-of-the-art pre-trained language model, T5-PEGASUS, to generate questions. Finally, the student can use this AI question directly or perform reflection by comparing his or her questions with the AI question. An empirical study was conducted to examine the quality of AI questions and the effect of this tool on student engagement and reading comprehension. The results of the study shows that students using this tool (treatment) were more engaged in generating low-level cognitive questions and performed better in acquiring knowledge than those using a traditional online question generation tool (control). These results indicate that student-AI question co-creation is beneficial to SQG training and educational assessment for reading comprehension, such as repeated practices.

28 October 2024
Verified
Conference Paper
Open access
JZJinxu Zhang
Jinxu Zhang
YYYongqi Yu
Yongqi Yu
YZYu Zhang
Yu Zhang
Document Visual Question Answering (DVQA) involves responding to queries based on the contents of document images. Existing works are confined to locating information within a single page and lack support for cross-page question-and-answer interactions. Furthermore, the token length limitation on model inputs can lead to the truncation of answer-relevant segments. In this study, we present CREAM, an innovative methodology that focuses on high-performance retrieval and integrates relevant multimodal document information to effectively address this critical issue. To overcome the limitations of current text embedding similarity methods, we first employ a coarse-to-fine retrieval and ranking approach. The coarse phase calculates the similarity between the query and text chunk embeddings, while the fine phase involves multiple rounds of grouping and ordering with a large language model to identify the text chunks most relevant to the query. Subsequently, integrating an attention pooling mechanism for multi-page document images into the vision encoder allows us to effectively merge the visual information of multi-page documents, enabling the multimodal large language model (MLLM) to simultaneously process both single-page and multi-page documents. Finally, we apply various parameter-efficient tuning methods to enhance document visual question-answering performance. Experiments demonstrate that our approach secures state-of-the-art results across various document datasets.