--- title: "Zhun Liu — author profile and papers" description: "Zhun Liu, Central South University: 2 papers on Elamur Research (2018–2022), h-index 25 — affiliations, topics, co-authors and who cites them in our…" canonical: "https://research.elamur.ai/free/en/authors/zhun-liu/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/zhun-liu/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Zhun Liu # Zhun Liu Zhun Liu's papers here, from 2018 to 2022, deal with the cost of building neural models that either combine several kinds of signal or grow to very large sizes. One line of work belongs to [vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/) and multimodal learning: how to merge separate language, audio and visual representations into a single joint vector without the expense exploding as more modalities are added. The other belongs to [language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), and specifically to the systems engineering needed to train a generative transformer too large for one GPU's memory. Both lines connect to [efficient models & compression](https://md.research.elamur.ai/free/en/topics/efficient-models/). The most cited work, [Efficient Low-rank Multimodal Fusion With Modality-Specific Factors](https://md.research.elamur.ai/free/en/papers/liu-2018-efficient-low-rank-multimodal-fusion/) (2018), begins from tensor-based fusion methods. Their outer-product tensors grow exponentially in size and computation with each extra modality, which also raises the risk of overfitting. The paper introduces Low-rank Multimodal Fusion, which decomposes the fusion weights into a few low-rank factors, one set per modality. Since the combined tensor itself breaks down into the unimodal vectors, the fused result can be obtained directly, without ever building the full high-dimensional object. The later paper, [Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B](https://md.research.elamur.ai/free/en/papers/smith-2022-using-deepspeed-megatron-train-megatron/) (2022), shifts from fusion to scale and to [text generation](https://md.research.elamur.ai/free/en/topics/text-generation/). It describes a joint Microsoft–NVIDIA project that trained an autoregressive, dense transformer of 530 billion parameters reading text left to right. The team joined Megatron-LM with DeepSpeed in a 3D-parallel design: model replicas sharded with ZeRO, a pipeline split across stages, and tensor slicing, with the tensor-parallel part confined to a single DGX node where NVLink bandwidth is high. Co-authors on these papers include [Louis-Philippe Morency](https://md.research.elamur.ai/free/en/authors/louis-philippe-morency/), [Paul Pu Liang](https://md.research.elamur.ai/free/en/authors/paul-pu-liang/), [Mostofa Patwary](https://md.research.elamur.ai/free/en/authors/mostofa-patwary/) and [Samyam Rajbhandari](https://md.research.elamur.ai/free/en/authors/samyam-rajbhandari/). Recent bylines list [Zhejiang University](https://md.research.elamur.ai/free/en/universities/zhejiang-university/) in Hangzhou. Zhun Liu has 2 papers in our library, in 2018–2022, most of them on [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), [Text generation](https://md.research.elamur.ai/free/en/topics/text-generation/), [Vision & language](https://md.research.elamur.ai/free/en/topics/multimodal/); often with [AmirAli Bagher Zadeh](https://md.research.elamur.ai/free/en/authors/amirali-bagher-zadeh/), [Jared Casper](https://md.research.elamur.ai/free/en/authors/jared-casper/), [Louis-Philippe Morency](https://md.research.elamur.ai/free/en/authors/louis-philippe-morency/). 2 papers in our library · 2018–2022 current affiliation [Central South University](https://md.research.elamur.ai/free/en/universities/central-south-university/), Changsha, China; [Northwestern Polytechnical University](https://md.research.elamur.ai/free/en/universities/northwestern-polytechnical-university/), Xi'an, China; [Zhejiang University](https://md.research.elamur.ai/free/en/universities/zhejiang-university/), Hangzhou, China [OpenAlex](https://openalex.org/A5045702830) · [ORCID](https://orcid.org/0000-0003-1455-8235) · [**h-index 25**](https://openalex.org/authors/A5045702830) · [121 works](https://openalex.org/works?filter=author.id:A5045702830) · [3 758 citations](https://openalex.org/works?filter=author.id:A5045702830&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [2 works in our library](https://md.research.elamur.ai/free/en/authors/zhun-liu/#papers) · [387 citations in our library](https://md.research.elamur.ai/free/en/authors/zhun-liu/cited-by/) · [⭐ Paper Star rank #1312](https://md.research.elamur.ai/free/en/paper-stars/page/14/#star-1312) 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=Zhun+Liu) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Zhun+Liu) · [dblp](https://dblp.org/search?q=Zhun+Liu) · [ACL Anthology](https://aclanthology.org/search/?q=Zhun+Liu) ## Affiliations - [Zhejiang University](https://md.research.elamur.ai/free/en/universities/zhejiang-university/) · Hangzhou, China —current · 2022–2026 on 6 papers - [Central South University](https://md.research.elamur.ai/free/en/universities/central-south-university/) · Changsha, China —current · 2025–2026 on 2 papers - [Northwestern Polytechnical University](https://md.research.elamur.ai/free/en/universities/northwestern-polytechnical-university/) · Xi'an, China —current · 2019–2026 on 2 papers - [Shaoxing University](https://ror.org/0435tej63) · Shaoxing, China —2021–2025 on 13 papers - [Huanghuai University](https://ror.org/02k92ks68) · Zhumadian, China —2019–2024 on 5 papers - [Dongguan University of Technology](https://ror.org/01m8p7q42) · Dongguan, China —2022–2023 on 4 papers - [Hebei North University](https://ror.org/03hqwnx39) · Zhangjiakou, China —2021–2022 on 2 papers - [Tan Kah Kee Innovation Laboratory](https://ror.org/05jxgts87) · Amoy, China —2021–2022 on 2 papers - [South China Normal University](https://ror.org/01kq0pv72) · Guangzhou, China —2020–2021 on 4 papers - [Jilin University](https://md.research.elamur.ai/free/en/universities/jilin-university/) · Changchun, China —2018–2020 on 7 papers From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 23 more institutions on single bylines are not listed. ## Topics [Language models & pretraining 1](https://md.research.elamur.ai/free/en/topics/language-models/) · [Text generation 1](https://md.research.elamur.ai/free/en/topics/text-generation/) · [Vision & language 1](https://md.research.elamur.ai/free/en/topics/multimodal/) · [Efficient models & compression 1](https://md.research.elamur.ai/free/en/topics/efficient-models/) ## Works by year ### 2020s [**2022** 1 paper](https://md.research.elamur.ai/free/en/papers/smith-2022-using-deepspeed-megatron-train-megatron/) ### 2010s [**2018** 1 paper](https://md.research.elamur.ai/free/en/papers/liu-2018-efficient-low-rank-multimodal-fusion/) ## Papers The most cited first (OpenAlex and our library). 1. **[Efficient Low-rank Multimodal Fusion With Modality-Specific Factors](https://md.research.elamur.ai/free/en/papers/liu-2018-efficient-low-rank-multimodal-fusion/)** *(2018 · 1 174 citations)* The paper tackles a scaling problem in multimodal fusion, which merges per-modality representations (language, audio, visual) into one joint vector. 2. **[Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model](https://md.research.elamur.ai/free/en/papers/smith-2022-using-deepspeed-megatron-train-megatron/)** *(2022 · 595 citations)* Large language models keep growing, and training them hits two limits. Their parameters no longer fit in one GPU's memory, and without careful co-design of software and hardware the training time becomes impractical. In a joint Microsoft–NVIDIA effort, the authors describe how they trained Megatron-Turing NLG 530B… ## Co-authors here 10 co-authors with a page, the most papers together first. - [AmirAli Bagher Zadeh](https://md.research.elamur.ai/free/en/authors/amirali-bagher-zadeh/) - [Jared Casper](https://md.research.elamur.ai/free/en/authors/jared-casper/) - [Louis-Philippe Morency](https://md.research.elamur.ai/free/en/authors/louis-philippe-morency/) - [Mostofa Patwary](https://md.research.elamur.ai/free/en/authors/mostofa-patwary/) - [Paul Pu Liang](https://md.research.elamur.ai/free/en/authors/paul-pu-liang/) - [Rewon Child](https://md.research.elamur.ai/free/en/authors/rewon-child/) - [Samyam Rajbhandari](https://md.research.elamur.ai/free/en/authors/samyam-rajbhandari/) - [Shrimai Prabhumoye](https://md.research.elamur.ai/free/en/authors/shrimai-prabhumoye/) - [Vijay Korthikanti](https://md.research.elamur.ai/free/en/authors/vijay-korthikanti/) - [Ying Shen](https://md.research.elamur.ai/free/en/authors/ying-shen/) ## Cited by Papers of our library cite their works 387 times (of the 140 884 papers whose bibliographies we read); [46 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/zhun-liu/cited-by/).