Jiahui GAO

Ph.D. candidate at HKU

github
linkedin
google

About Me

Hi! I am a fourth-year Ph.D. candidate of Department of Statistics and Actuarial Science (SAAS) in the University of Hong Kong (HKU), supervised by Stephen M.S. LEE and Lingpeng Kong. During the first year of my Ph.D. studies and my Master's studies in HKU, I was fortunate to be supervised by Yu Leung Ho Philip.

My current research interest is Natural Language Processing, including Pre-trained Language Modeling (PLM), Automatic Machining Learning (AutoML), and Multi-Modal (vision-language) learning.

News

[Jan 2023] Two of our papers are accpted by ICLR-2023 as spotlight papers.

[Dec 2022] Attended EMNLP 2022 at Abu Dhabi.

[Oct 2022] Two of our papers are accpted by EMNLP-2022.

[Jan 2022] Our paper is accepted by ICLR-2022 as spotlight paper.

[Dec 2021] Two of our papers are accepted by AAAI-2022, including one oral paper.

[May 2021] Our paper is accepted by ICML-2021.

Research

(* indicates equal contribution)

DetGPT: Detect What You Need via Reasoning
R.Pi*, Jiahui Gao*, S. Diao*, R. Pan, H. Dong, J. Zhang, L. Yao, J. Han, H. Xu, L. Kong, T. Zhang
Preprint. [arxiv]

SunGen: Self-Guided High-Quality Data Generation in Efficient Zero-Shot Learning
Jiahui Gao*, R. Pi *, Y. Lin, H. Xu, J. Ye, Z. Wu, W. Zhang, X. Liang, Z. Li, L. Kong.
International Conference on Learning Representations (ICLR-2023). [arxiv]
(Spotlight paper)

A Holistic View of Noise Transition Matrix in Deep Learning and Beyond?
Y. Lin, R. Pi, W. Zhang, X. Xia, Jiahui Gao, X. Zhou, T. Liu
International Conference on Learning Representations (ICLR-2023). [arxiv]
(Spotlight paper)

ZeroGen: Efficient Zero-shot Learning via Dataset Generation
J. Ye*, Jiahui Gao*, Q. Li, H. Xu, J. Feng, Z. Wu, T. Yu, and L. Kong.
Conference on Empirical Methods in Natural Language Processing (EMNLP-2022).[arxiv] [code]

ProGen: Progressive Zero-shot Dataset Generation via In-context Feedback
J. Ye, Jiahui Gao, Z. Wu, J. Feng, T. Yu, and L. Kong.
Findings of the Conference on Empirical Methods in Natural Language Processing (EMNLP-2022 Findings).[arxiv] [code]

Revisiting Over-smoothing in BERT from the Perspective of Graph
H. Shi*, Jiahui Gao*, H. Xu, X. Liang, Z. Li, L. Kong, S.M.S. Lee, J. T. Kwok.
International Conference on Learning Representations (ICLR-2022). [arxiv]
(Spotlight paper, rate: 5.19%)

AutoBERT-Zero: Evolving BERT Backbone from Scratch
Jiahui Gao, H. Xu, H. Shi, X. Ren, P.L.H. Yu, X. Liang, X. Jiang, Z. Li.
Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-2022). [arxiv]

UNISON: Unpaired Cross-Lingual Image Captioning
Jiahui Gao, Z. Yi, P.L.H. Yu, S. Joty, and J. Gu.
Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-2022). [arxiv][code]
(Oral paper)

SparseBERT: Rethinking the Importance Analysis in Self-Attention
S. Han, Jiahui Gao, X. Ren, H. Xu, X. Liang, Z. Li, and J. T. Kwok.
International Conference on Machine Learning (ICML-2021). [arxiv] [code]

Detecting Comments Showing Risk for Suicide in YouTube
Jiahui Gao, Q. Cheng and P.L.H. Yu.
Proceedings of the Future Technologies Conference (FTC-2018). [arxiv]

Recent Talks

Experiences

  • Remote research cooperation with Dr. Jiuxiang Gu from Adobe Research, Dec. 2019 - Aug. 2020.
  • Business Analyst (Full-time), Alibaba, Hangzhou, Aug. 2018 - Aug. 2019.
  • Business Analyst (Summer Intern), Tencent, Shenzhen, Jun. 2017 - Aug. 2017.

Awards & Honors

Teaching

  • STAT7102/7614: Advanced Statistical Modelling
  • ARIN7102: Applied Data Mining and Text Analytics
  • STAT2604: Introduction to R/Python Programming and Elementary Data Analysis