[Google Scholar] [Twitter] [Github]

News

[2024.10]. 1 paper accepted to TMLR.

[2024.9]. 1 paper accepted to EMNLP 2024.

[2024.6]. Serve as Area Chairs/Action Editors for ACL, EMNLP, ARR.

[2024.5]. 1 paper accepted to ICML 2024.

[2024.3]. 2 papers accepted to NAACL 2024 (1 outstanding paper).

[2024.2]. 1 paper accepted to CVPR 2024.

[2024.1]. 2 papers accepted to ICLR 2024.

[2023]. ……

Bio

I’m a third-year CS Ph.D. student at UIUC. I’m advised by Prof. Heng Ji and I also collaborate with Prof. Hao Peng. I work on scalable AI models, aiming to establish fundamental approaches to address the following challenges:

• Predictable Scaling: The scaling laws that can reliably predict the development of foundation models.
• Scalable Modeling: The model architectures and training recipes for multimodality that can transform increasing compute and data into sustained performance gains without hitting any bottleneck.
• Scalable Oversight: The scalable post-training approaches that can continually improve the foundation models and align their objectives with human intents even as they exceed human capabilities.

In my undergraduate years, I was a research intern at THUNLP advised by Prof. Zhiyuan Liu. In the early stages of my research, I worked closely with Dr. Fanchao Qi and received great help from him. Also, I am delighted to work with Prof. Wei Wei and Prof. Dawn Song.

Selected Publications

[Full Publications]

^* indicates equal contribution ⁺ indicates corresponding author

• Scaling Laws for Predicting Downstream Performance in LLMs [paper]
  Yangyi Chen, Binxuan Huang, Yifan Gao, Zhengyang Wang, Jingfeng Yang, Heng Ji
  Arxiv

• A Single Transformer for Scalable Vision-Language Modeling [paper]
  Yangyi Chen^*, Xingyao Wang^*, Hao Peng, Heng Ji
  TMLR 2024

• SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales [paper]
  Tianyang Xu^*, Shujin Wu^*, Shizhe Diao, Xiaoze Liu, Xingyao Wang, Yangyi Chen⁺, Jing Gao⁺
  EMNLP 2024

• DRESS: Instructing Large Vision-Language Models to Align and Interact with Humans via Natural Language Feedback [paper]
  Yangyi Chen, Karan Sikka, Michael Cogswell, Heng Ji, Ajay Divakaran
  CVPR 2024

• A Close Look into the Calibration of Pre-trained Language Models [paper]
  Yangyi Chen^*, Lifan Yuan^*, Ganqu Cui, Zhiyuan Liu, Heng Ji.
  ACL 2023

Education

• University of Illinois Urbana-Champaign, Ph.D. in Computer Science; 2022-Present

• Huazhong University of Science and Technology, BS in Software Engineering; 2018-2022

Experience

• Amazon@Palo Alto, Applied Scientist Research Intern; May-Aug 2024; Hosted by Dr. Binxuan Huang.

• SRI International@New Jersey, Research Intern; May-Aug 2023; Hosted by Dr. Karan Sikka, Dr. Michael Cogswell, and Dr. Ajay Divakaran.

Service

Area Chair/Action Editor

2024: ACL/ARR (Feb), ARR (Apr), EMNLP/ARR (Jun)

Reviewer

2024: COLM, ICML

2023: ICLR, EMNLP, EMNLP Industry Track, NeurIPS, NeurIPS D&B Track, ARR (Feb, Apr, Jun, Aug, Oct, Dec), ACL (Outstanding Award), IEEE T-IFS

2022: NeurIPS D&B Track, EMNLP, ARR (Dec)

Assistant: AAAI 2022, ARR 2022 (Jan), EMNLP 2021, ARR 2021 (Oct, Nov)

Personal

When I was young (around 20), I enjoyed playing the piano and basketball. When I started doing research, my interests changed and I quickly fell in love with food and soap operas. But always, I am a big fan of milk tea!!