--- title: "Jiazheng Xu — author profile and papers" description: "Jiazheng Xu, Harbin Institute of Technology: 4 papers on Elamur Research (2023–2024), h-index 7 — affiliations, topics, co-authors and who cites them in…" canonical: "https://research.elamur.ai/free/en/authors/jiazheng-xu/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/jiazheng-xu/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Jiazheng Xu # Jiazheng Xu Jiazheng Xu works on large multimodal and language models. The papers here, from 2023 and 2024, cover four problems: joining pretrained language models to vision, building agents that act on graphical interfaces from screenshots, generating video from text prompts, and testing how well models reason over very long inputs. A recurring design choice is to add modality-specific "expert" components inside a transformer, so the model gains new abilities without losing what it already does well. The most cited of these works is [CogVLM](https://md.research.elamur.ai/free/en/papers/wang-2023-cogvlm-visual-expert-pretrained/) (2023). It leaves the pretrained language model untouched and inserts trainable visual expert weights into each layer, which only image tokens pass through. The aim is a deeper fusion of vision and language than a small adapter gives, while avoiding the loss of text skills that full fine-tuning causes. [CogAgent](https://md.research.elamur.ai/free/en/papers/hong-2023-cogagent-visual-language-model/) (2023) builds on CogVLM-17B to make an 18B-parameter model for GUI agents. It adds a lightweight high-resolution image branch that feeds into every decoder layer through cross-attention, so the model can read small on-screen text and fine layout details. The newer papers move toward video generation and evaluation. [CogVideoX](https://md.research.elamur.ai/free/en/papers/yang-2024-cogvideox-text-video-diffusion/) (2024) applies a diffusion transformer with expert components to text-to-video synthesis. It targets clips that are longer, show more motion and follow prompts describing a series of events. [LongBench v2](https://md.research.elamur.ai/free/en/papers/bai-2024-longbench-v2-deeper-understanding/) (2024) is a benchmark of 503 multiple-choice questions written by human annotators, with contexts ranging from 8k to 2M words. Each question was kept only if it could not be solved by several long-context models or by an expert reviewer working under a time limit. Xu is currently at [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/) and was earlier at [Tsinghua University](https://md.research.elamur.ai/free/en/universities/tsinghua-university/), from 2022 to 2025. The most frequent co-author here is [Wenyi Hong](https://md.research.elamur.ai/free/en/authors/wenyi-hong/), and other collaborators include [Yuxiao Dong](https://md.research.elamur.ai/free/en/authors/yuxiao-dong/) and [Juanzi Li](https://md.research.elamur.ai/free/en/authors/juanzi-li/). The papers fall under [Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/), language models and pretraining, and reasoning. Jiazheng Xu has 4 papers in our library, in 2023–2024, most of them on [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), [Reasoning](https://md.research.elamur.ai/free/en/topics/reasoning/), [Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/); often with [Wenyi Hong](https://md.research.elamur.ai/free/en/authors/wenyi-hong/), [Bin Xu](https://md.research.elamur.ai/free/en/authors/bin-xu/), [Juanzi Li](https://md.research.elamur.ai/free/en/authors/juanzi-li/). 4 papers in our library · 2023–2024 current affiliation [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/), Harbin, China [OpenAlex](https://openalex.org/A5042981778) · [**h-index 7**](https://openalex.org/authors/A5042981778) · [19 works](https://openalex.org/works?filter=author.id:A5042981778) · [418 citations](https://openalex.org/works?filter=author.id:A5042981778&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [4 works in our library](https://md.research.elamur.ai/free/en/authors/jiazheng-xu/#papers) · [653 citations in our library](https://md.research.elamur.ai/free/en/authors/jiazheng-xu/cited-by/) · [⭐ Paper Star rank #1967](https://md.research.elamur.ai/free/en/paper-stars/page/20/#star-1967) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Jiazheng+Xu) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Jiazheng+Xu) · [dblp](https://dblp.org/search?q=Jiazheng+Xu) · [ACL Anthology](https://aclanthology.org/search/?q=Jiazheng+Xu) ## Affiliations - [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/) · Harbin, China —current · 2025–2026 on 5 papers - [Tsinghua University](https://md.research.elamur.ai/free/en/universities/tsinghua-university/) · Beijing, China —2022–2025 on 6 papers From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 2 more institutions on single bylines are not listed. ## Topics [Language models & pretraining 1](https://md.research.elamur.ai/free/en/topics/language-models/) · [Reasoning 1](https://md.research.elamur.ai/free/en/topics/reasoning/) · [Vision & language 1](https://md.research.elamur.ai/free/en/topics/multimodal/) ## Works by year ### 2020s [**2024** 2 papers](https://md.research.elamur.ai/free/en/authors/jiazheng-xu/year/2024/) · [**2023** 2 papers](https://md.research.elamur.ai/free/en/authors/jiazheng-xu/year/2023/) ## Papers The most cited first (OpenAlex and our library). 1. **[CogVLM: Visual Expert for Pretrained Language Models](https://md.research.elamur.ai/free/en/papers/wang-2023-cogvlm-visual-expert-pretrained/)** *(2023 · 373 citations)* Many vision-language models attach a frozen image encoder to a frozen LLM through a small adapter. This "shallow alignment" treats image features as extra input tokens, and the authors argue it limits how deeply vision and language interact. The alternative, fine-tuning the whole LLM on image-text data, erodes its… 2. **[CogAgent: A Visual Language Model for GUI Agents](https://md.research.elamur.ai/free/en/papers/hong-2023-cogagent-visual-language-model/)** *(2023 · 353 citations)* Most interactions with computers and phones happen through graphical interfaces. Text-only LLM agents handle these poorly because GUIs often lack APIs, carry meaning through icons and layout, and include elements that HTML cannot describe. The authors present CogAgent, an 18B-parameter visual language model built on… 3. **[CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer](https://md.research.elamur.ai/free/en/papers/yang-2024-cogvideox-text-video-diffusion/)** *(2024 · 163 citations)* CogVideoX addresses a persistent weakness of text-to-video systems: clips that are short, nearly static, and poorly matched to prompts describing a sequence of events. The authors build a diffusion transformer around four main components: 4. **[LongBench v2: Towards Deeper Understanding and Reasoning on Realistic Long-context Multitasks](https://md.research.elamur.ai/free/en/papers/bai-2024-longbench-v2-deeper-understanding/)** *(2024 · 130 citations)* Most long-context benchmarks reward shallow skills. Answers can often be pulled straight from the text, tasks are frequently synthetic, and scoring relies on fragile metrics like F1 or ROUGE. LongBench v2 targets genuine comprehension and reasoning over very long inputs instead. The authors gathered 503 English… ## Co-authors here 12 of 26 co-authors with a page, the most papers together first. - [Wenyi Hong](https://md.research.elamur.ai/free/en/authors/wenyi-hong/) ×3 - [Bin Xu](https://md.research.elamur.ai/free/en/authors/bin-xu/) ×2 - [Juanzi Li](https://md.research.elamur.ai/free/en/authors/juanzi-li/) ×2 - [Junhui Ji](https://md.research.elamur.ai/free/en/authors/junhui-ji/) ×2 - [Qingsong Lv](https://md.research.elamur.ai/free/en/authors/qingsong-lv/) ×2 - [Weihan Wang](https://md.research.elamur.ai/free/en/authors/weihan-wang/) ×2 - [Wenmeng Yu](https://md.research.elamur.ai/free/en/authors/wenmeng-yu/) ×2 - [Yan Wang](https://md.research.elamur.ai/free/en/authors/yan-wang/) ×2 - [Yuxiao Dong](https://md.research.elamur.ai/free/en/authors/yuxiao-dong/) ×2 - [Yuxuan Zhang](https://md.research.elamur.ai/free/en/authors/yuxuan-zhang/) ×2 - [Zhuoyi Yang](https://md.research.elamur.ai/free/en/authors/zhuoyi-yang/) ×2 - [Da Yin](https://md.research.elamur.ai/free/en/authors/da-yin/) ## Cited by Papers of our library cite their works 653 times (of the 140 884 papers whose bibliographies we read); [36 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/jiazheng-xu/cited-by/).