[about]   [papers]   [travels]  

my face

Chenglei Si

[email] [scholar] [twitter] [github]
1st Year PhD Student at Stanford NLP


At Stanford, I'm rotating with Diyi Yang and Michael Bernstein. Before coming to Stanford, I did my undergrad at the University of Maryland where I was advised by Jordan Boyd-Graber, while also working closely with Hal Daumé III, He He, Danqi Chen, and Sherry Wu. In summer 2022, I did a research internship at Microsoft hosted by Zhe Gan. Before that, I got into NLP research by working with Min-Yen Kan and Zhiyuan Liu.

Nowadays, I'm fascinated by Human-Muppet Interaction. And I'm particularly concerned about the following questions:
How can humans verify what Muppets said, especially in tasks where humans lack the domain expertise?
How can we enable human-Muppet collaboration to complete tasks that humans or Muppets cannot solve alone?
What is the long-term impact of humans relying on Muppets?
How can we measure and improve the safety of Muppets?
In pursuing these research questions, I tend to move away from existing benchmarks and put human needs at the center of my research. I also aim to craft an interdisciplinary research agenda that connects insights from HCI, NLP, ML, Psychology, and Linguistics.

Back in the old days, I also worked on Question Answering, Tokenization, and Prompting.

Recent Papers

  • Large Language Models Help Humans Verify Truthfulness -- Except When They Are Convincingly Wrong
    Chenglei Si, Navita Goyal, Sherry Tongshuang Wu, Chen Zhao, Shi Feng, Hal Daumé III, Jordan Boyd-Graber
    preprint [paper] [tweet]

  • Mixture of Prompt Experts for Generalizable and Interpretable Question Answering
    Chenglei Si, Weijia Shi, Chen Zhao, Luke Zettlemoyer, Jordan Boyd-Graber
    EMNLP 2023 Findings [paper] [code]

  • Ignore This Title and HackAPrompt: Exposing Systemic Vulnerabilities of LLMs through a Global Scale Prompt Hacking Competition
    Sander Schulhoff*, Jeremy Pinto*, Anaum Khan, Louis-François Bouchard, Chenglei Si, Svetlina Anati, Valen Tagliabue, Anson Liu Kost, Christopher Carnahan, Jordan Boyd-Graber
    EMNLP 2023 [webpage]

  • Measuring Inductive Biases of In-Context Learning with Underspecified Demonstrations
    Chenglei Si*, Dan Friedman*, Nitish Joshi, Shi Feng, Danqi Chen, He He
    ACL 2023 [paper] [code] [tweet] [OpenReview]

  • Prompting GPT-3 To Be Reliable
    Chenglei Si, Zhe Gan, Zhengyuan Yang, Shuohang Wang, Jianfeng Wang, Jordan Boyd-Graber, Lijuan Wang
    ICLR 2023 [paper] [code] [tweet] [video]

  • Sub-Character Tokenization for Chinese Pretrained Language Models
    Chenglei Si*, Zhengyan Zhang*, Yingfa Chen*, Fanchao Qi, Xiaozhi Wang, Zhiyuan Liu, Yasheng Wang, Qun Liu, Maosong Sun
    TACL 2023 [paper] [code]

  • Re-Examining Calibration: The Case of Question Answering
    Chenglei Si, Chen Zhao, Sewon Min, Jordan Boyd-Graber
    EMNLP 2022 Findings [paper] [code] [video]


  • Oct 2023, UIST @ San Francisco
  • July 2023, ACL @ Toronto
  • May 2023, ICLR @ Kigali
  • March 2023, Visit Day @ Stanford
  • March 2023, Visit Day @ NYU
  • Dec 2022, EMNLP @ Abu Dhabi
  • Summer 2022, Internship + NAACL @ Seattle