# Korawat Tanwisuth

### Featured Co-authors

- [Mingyuan Zhou](/content/profile/mingyuan-zhou/index.html) - 110 publications
- [Bo Chen](/content/profile/bo-chen/index.html) - 105 publications
- [Pengcheng He](/content/profile/pengcheng-he/index.html) - 53 publications
- [He Zhao](/content/profile/he-zhao/index.html) - 39 publications
- [Xiaoning Qian](/content/profile/xiaoning-qian/index.html) - 38 publications
- [Dongsheng Wang](/content/profile/dongsheng-wang/index.html) - 22 publications
- [Huangjie Zheng](/content/profile/huangjie-zheng/index.html) - 22 publications
- [Shujian Zhang](/content/profile/shujian-zhang/index.html) - 17 publications
- [Dandan Guo](/content/profile/dandan-guo/index.html) - 9 publications
- [Xinjie Fan](/content/profile/xinjie-fan/index.html) - 8 publications

## Research

### [POUF: Prompt-oriented unsupervised fine-tuning for large pre-trained models](/content/publication/pouf-prompt-oriented-unsupervised-fine-tuning-for-large-pre-trained-models/index.html)
Through prompting, large-scale pre-trained models have become more expressive.

### [A Prototype-Oriented Clustering for Domain Shift with Source Privacy](/content/publication/a-prototype-oriented-clustering-for-domain-shift-with-source-privacy/index.html)
Unsupervised clustering under domain shift (UCDS) studies how to transfer knowledge effectively.

### [Representing Mixtures of Word Embeddings with Mixtures of Topic Embeddings](/content/publication/representing-mixtures-of-word-embeddings-with-mixtures-of-topic-embeddings/index.html)
A topic model is often formulated as a generative model that explains how to represent data.

### [Alignment Attention by Matching Key and Query Distributions](/content/publication/alignment-attention-by-matching-key-and-query-distributions/index.html)
The neural attention mechanism has been incorporated into deep neural networks.

### [Contextual Dropout: An Efficient Sample-Dependent Dropout Module](/content/publication/contextual-dropout-an-efficient-sample-dependent-dropout-module/index.html)
Dropout has been demonstrated as a simple and effective module.