--- title: "Xiaohua Zhai — author profile and papers" description: "Xiaohua Zhai, Google: 2 papers on Elamur Research (2020–2023), h-index 35 — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/xiaohua-zhai/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/xiaohua-zhai/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Xiaohua Zhai # Xiaohua Zhai Xiaohua Zhai works on large-scale visual representation learning, with papers here centred on two linked questions: whether architectures built for language can serve image recognition with little change, and how images and text can be trained together efficiently at scale. The work typically treats engineering constraints, such as memory and communication across many devices, as part of the research problem. It sits within the wider area of [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/). The most cited of Zhai's papers here is [An Image is Worth 16x16 Words](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/) (2020), which introduced the Vision Transformer. The model cuts an image into fixed-size patches, projects each into a vector and feeds the resulting sequence to an almost unmodified Transformer encoder. It asks whether convolutional networks are really needed for classification, and it notes that without vision-specific assumptions such as locality the model lags a comparable ResNet when trained on ImageNet alone. The second paper, [Sigmoid Loss for Language Image Pre-Training](https://md.research.elamur.ai/free/en/papers/zhai-2023-sigmoid-loss-language-image-pre/) (2023), revisits contrastive image-text training of the CLIP and ALIGN kind. It swaps the batch-wide softmax for a pairwise sigmoid objective that treats every image-text pair as its own binary decision. The 2023 sigmoid-loss paper shows where the work has moved: from image-only models towards joint vision-language pre-training. Because the new loss needs no global normalisation, it can be computed in small blocks spread across devices, and this keeps memory use low even for large batches. Zhai is currently at [Google](https://md.research.elamur.ai/free/en/companies/google/) and [Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/), and earlier worked at [Google in Zurich](https://md.research.elamur.ai/free/en/companies/google-switzerland/). Frequent co-authors include [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/) and [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/), along with collaborators such as Mostafa Dehghani, Thomas Unterthiner, Jakob Uszkoreit and Basil Mustafa. Xiaohua Zhai has 2 papers in our library, in 2020–2023, most of them on [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/); often with [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/), [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/), [Basil Mustafa](https://md.research.elamur.ai/free/en/authors/basil-mustafa/). 2 papers in our library · 2020–2023 [OpenAlex](https://openalex.org/A5071668416) · [**h-index 35**](https://openalex.org/authors/A5071668416) · [73 works](https://openalex.org/works?filter=author.id:A5071668416) · [31,295 citations](https://openalex.org/works?filter=author.id:A5071668416&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [📄 2 works in our library](https://md.research.elamur.ai/free/en/authors/xiaohua-zhai/#papers) · [🔗 2,648 citations in our library](https://md.research.elamur.ai/free/en/authors/xiaohua-zhai/cited-by/) · [2 more in our library without a page yet](#also) · [⭐ Paper Star rank #762](https://md.research.elamur.ai/free/en/paper-stars/page/8/#star-762) 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=Xiaohua+Zhai) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Xiaohua+Zhai) · [dblp](https://dblp.org/search?q=Xiaohua+Zhai) · [ACL Anthology](https://aclanthology.org/search/?q=Xiaohua+Zhai) ## Affiliations [🏢**Google** Mountain View · United States 2018–2025 · 29 papers*current*](https://md.research.elamur.ai/free/en/companies/google/) · [🏢**Google DeepMind** London · United Kingdom 2023–2025 · 6 papers*current*](https://md.research.elamur.ai/free/en/companies/google-deepmind/) · [🏢**Google** Zurich · Switzerland 2019–2024 · 12 papers](https://md.research.elamur.ai/free/en/companies/google-switzerland/) · [🎓**Peking University** Beijing · China 2012–2015 · 8 papers](https://md.research.elamur.ai/free/en/universities/peking-university/) From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 1 more institution on single bylines are not listed. ## Topics [Language models & pretraining (1)](https://md.research.elamur.ai/free/en/topics/language-models/) ## Works by year ### 2020s [**2023** 1 paper](https://md.research.elamur.ai/free/en/papers/zhai-2023-sigmoid-loss-language-image-pre/) · [**2020** 1 paper](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/) ## Papers The most cited first (OpenAlex and our library). 1. **[An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://md.research.elamur.ai/free/en/papers/dosovitskiy-2020-image-worth-16x16-words-transformers/)** *(2020 · 1,947 citations)* The authors from Google Research test whether convolutional networks are truly necessary for image classification. 2. **[Sigmoid Loss for Language Image Pre-Training](https://md.research.elamur.ai/free/en/papers/zhai-2023-sigmoid-loss-language-image-pre/)** *(2023 · 1,427 citations)* Contrastive image-text pre-training methods such as CLIP and ALIGN usually use a softmax loss, which normalizes similarity scores over the whole batch. ## Also in our library 2 more works of theirs are in our library without a page here yet; they will appear as we publish them. ## Co-authors here 6 co-authors with a page, the most papers together first. - [Alexander Kolesnikov](https://md.research.elamur.ai/free/en/authors/alexander-kolesnikov/) ×2 - [Lucas Beyer](https://md.research.elamur.ai/free/en/authors/lucas-beyer/) ×2 - [Basil Mustafa](https://md.research.elamur.ai/free/en/authors/basil-mustafa/) - [Jakob Uszkoreit](https://md.research.elamur.ai/free/en/authors/jakob-uszkoreit/) - [Mostafa Dehghani](https://md.research.elamur.ai/free/en/authors/mostafa-dehghani/) - [Thomas Unterthiner](https://md.research.elamur.ai/free/en/authors/thomas-unterthiner/) ## Cited by Papers of our library cite their works 2,648 times (of the 140,884 papers whose bibliographies we read); [98 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/xiaohua-zhai/cited-by/).