Weiyan Shi

Welcome! I am a Postdoc at Stanford NLP, working with Diyi Yang. I graduated with my PhD in Computer Science at Columbia University, advised by Prof. Zhou Yu, working on Natural Language Processing. Previously, I completed my master’s degree in Statistics from the University of California, Berkeley and my B.S. in Mathematics from Renmin University of China. I have worked in the SF bay area as a full-time data scientist on customer service chatbots for two years before starting my PhD.

I am joining Northeastern University as a tenure-track assistant professor in Fall 2024. If you are interested in working with me, make sure to fill out this form (PhD/intern) or this form (Postdoc) and send me an email :)

Email: weiyans@stanford.edu

Google Scholar / CV

Links: Research Overview / Updates / Awards / Papers / Teaching / Services / Miscellaneous

Research Overview

My research interests are in Natural Language Processing, especially intelligent interactive systems 🤖 and the following directions:

• Interactive systems specialized in social influence for social good (e.g., persuasive dialogues): [ACL 2019], [CHI 2020], [AAAI 2020], [EMNLP 2020a], [EMNLP 2020b], [EMNLP 2021], [AACL 2022], [New preprint], [Science]
• Privacy-preserving NLP models: [NAACL 2021], [EMNLP 2022]
• Task-oriented and open-domain dialogue systems: [ACL 2018], [NAACL 2019], [EMNLP 2019], [EMNLP 2020], [ACL 2021], [New preprint]
• Intelligible dialogue generation: [EMNLP 2021] [New preprint]
• Learning through interaction: [New preprint]

My research vision is to build a natural interface between human intelligence and machine intelligence via natural conversations, so that all members of society can interact with AI models seamlessly regardless of their backgrounds.

Updates

Jan 2024: Checkout our new papers on how to persuade LLMs to jailbreak them with a success rate of 92%, and how to persuade them to believe in misinformation!

July 2023: Checkout CHATS’ lab’s first project, KokoMind

2023: Talk “Interactive AI Systems Specialized in Social Influence” at the University of Hawaii (2023/01), Rice (2023/01), Northwestern Stats and Data Science (2023/01), ASU (2023/02), Purdue (2023/02), CMU (2023/02), Northeastern (2023/03), UW-Madison i-School (2023/03), PSU (2023/03), Cornell Tech (2023/04), NYU (2023/05), Stanford NLP (2023/06).

Jan 2023: Our Cicero work made it into the New York Times front page!

Nov 2022: Check out our Science publication on the first human-level negotiation AI agent that can negotiate, persuade, and coordinate with human players in natural language in the classic complex 7-player board game of Diplomacy! This tackles several challenges such as multi-party dialogue modeling, strategic reasoning, social influence and how to connect them in intense and length conversations! [Meta AI blog post] [code] [Commentary in Science News]

Oct 2022: I am excited to be named as Rising Stars in Machine Learning, 2022! Thanks UMD :)

Oct 2022: Our paper on privacy-preserving Transformer-based models is accepted to EMNLP 2022.

Sep 2022: I am co-teaching Conversational AI (COMS 6998) with Zhou at Columbia.

Jun - Sep 2022: This summer I interned with Jason Weston and Jing Xu to incorporate human feedback to continuously improve deployed chatbots (checkout our paper).

Jun - Dec 2021: Finished my internship with Mike Lewis on zero-shot dialogue nonsense detection.

Jun - Dec 2020: Finished my internship with Mike Lewis on strategic dialogue.

Awards

• Rising Stars in AI, KAUST, 2023
• Rising Stars in EECS, 2023
• Rising Stars in Machine Learning, 2022
• Best Paper Nomination, ACL 2019
• Dean’s Distinguished PhD Fellowship
• Department Citation (top 1), UC Berkeley, 2016
• Speaker at Department Commencement (top 1), UC Berkeley, 2016
• National Scholarship, 2014
• Presidential Fellowship for Studying Abroad, 2013

Papers

Google Scholar

2024

• How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs
  Yi Zeng*, Hongpeng Lin* (lead authors), Jingwen Zhang, Diyi Yang, Ruoxi Jia*, Weiyan Shi* (co-supervise)
  arXiv, project page, code and data
  

2023

• The Earth is Flat because…: Investigating LLMs’ Belief towards Misinformation via Persuasive Conversation
  Rongwu Xu, Brian S. Lin, Shujian Yang, Tianqi Zhang, Weiyan Shi*, Tianwei Zhang, Zhixuan Fang, Wei Xu, Han Qiu* (equal advising)
  arXiv, project page, code and data
  

• AutoReply: Detecting Nonsense in Dialogue Introspectively with Discriminative Replies
  Weiyan Shi, Emily Dinan, Adi Renduchintala, Daniel Fried, Athul Paul Jacob, Zhou Yu, Mike Lewis
  EMNLP Findings, 2023
  

• Social Influence Dialogue Systems: A Scoping Survey of the Efforts Towards Influence Capabilities of Dialogue Systems
  Kushal Chawla*, Weiyan Shi*(equal contribution), Jingwen Zhang**, Gale Lucas**, Zhou Yu**, Jonathan Gratch**(co-supervision)
  EACL, 2023
  

2022

• Human-Level Play in the Game of Diplomacy by Combining Language Models with Strategic Reasoning
  FAIR, Anton Bakhtin*, Noam Brown*, Emily Dinan*, Gabriele Farina, Colin Flaherty*, Daniel Fried, Andrew Goff, Jonathan Gray*, Hengyuan Hu*, Athul Paul Jacob*, Mojtaba Komeili, Karthik Konath, Minae Kwon, Adam Lerer*, Mike Lewis*, Alexander H. Miller*, Sasha Mitts, Adithya Renduchintala*, Stephen Roller, Dirk Rowe, Weiyan Shi*, Joe Spisak, Alexander Wei, David Wu*, Hugh Zhang*, Markus Zijlstra
  (*A member of the core research team. Authors listed alphabetically.)
  Science 2022, [Meta AI blog post], [code]
  Commentary in Science News by Matthew Hutson, Yoram Bachrach, Noam Brown, Jonathan Gratch, and Zhou Yu
  New York Times (front page)
  Selected media coverage: New York Times, The Washington Post, The Economist, MIT Technology Review, Forbes
  

• Just Fine-tune Twice: Selective Differential Privacy for Large Language Models
  Weiyan Shi, Ryan Shea, Si Chen, Chiyuan Zhang, Ruoxi Jia, Zhou Yu
  EMNLP 2022, code and data
  

• Seamlessly Integrating Factual Information and Social Content with Persuasive Dialogue
  Maximillian Chen, Weiyan Shi, Feifan Yan, Ryan Hou, Jingwen Zhang, Saurav Sahay, Zhou Yu
  AACL 2022
  

• Selective Differential Privacy for Language Modeling
  Weiyan Shi, Aiqi Cui, Evan Li, Ruoxi Jia, Zhou Yu
  NAACL 2022, code and data, talk
  

2021

• Refine and Imitate: Reducing Repetition and Inconsistency in Persuasion Dialogues via Reinforcement Learning and Human Demonstration
  Weiyan Shi, Yu Li, Saurav Sahay, Zhou Yu
  EMNLP 2021 Findings, code
  

• LEGOEval: An Open-Source Toolkit for Dialogue System Evaluation via Crowdsourcing
  Yu Li, Josh Arnold, Feifan Yan, Weiyan Shi, Zhou Yu
  ACL 2021 Demo, code, demo
  

• PRAL: A tailored pre-training model for task-oriented dialogue generation
  Jing Gu, Qingyang Wu, Chongruo Wu, Weiyan Shi, Zhou Yu
  ACL 2021, talk
  

2020

• INSPIRED: Toward Sociable Recommendation Dialogue Systems
  Shirley Anugrah Hayati, Dongyeop Kang, Qingxiaoyang Zhu, Weiyan Shi, Zhou Yu
  EMNLP 2020, code and data, talk
  

• Structured Attention for Unsupervised Dialogue Structure Induction
  Liang Qiu, Yizhou Zhao, Weiyan Shi, Yuan Liang, Feng Shi, Tao Yuan, Zhou Yu, Song-Chun Zhu
  EMNLP 2020, code, talk
  

• Understanding User Resistance Strategies in Persuasive Conversations
  Youzhi Tian, Weiyan Shi, Chen Li, Zhou Yu
  EMNLP 2020 Findings
  

• Effects of Persuasive Dialogues: Testing Bot Identities and Inquiry Strategies
  Weiyan Shi, Xuewei Wang,Yoo Jung Oh, Jingwen Zhang, Saurav Sahay, Zhou Yu
  CHI 2020, talk
  

• End-to-End Trainable Non-Collaborative Dialogue Systems
  Yu Li, Kun Qian, Weiyan Shi, Zhou Yu
  AAAI 2020, code
  

2019

• How to Build User Simulators to Train RL-based Dialogue Systems
  Weiyan Shi*, Kun Qian(equal contribution), Xuewei Wang, Zhou Yu
  *EMNLP 2019, code
  

• Persuasion for Good: Towards a Personalized Persuasive Dialogue System for Social Good
  Xuewei Wang*, Weiyan Shi*(equal contribution), Richard Kim, Yoojung Oh, Sijia Yang, Jingwen Zhang, Zhou Yu
  ACL 2019, code and data, talk, Best Paper Nomination
  

• Unsupervised Dialogue Structure Learning
  Weiyan Shi, Tiancheng Zhao, Zhou Yu
  NAACL 2019, code
  

2018

• Sentiment Adaptive End-to-End Dialogue Systems
  Weiyan Shi, Zhou Yu
  ACL 2018, data
  

Preprints

• When Life Gives You Lemons, Make Cherryade: Converting Feedback from Bad Responses into Good Labels
  Weiyan Shi, Emily Dinan, Kurt Shuster, Jason Weston*, Jing Xu*(equal contribution)
  arXiv, 2022
  

Teaching

• Co-instructor, Conversational AI Special Topics (COMS 6998)
  Columbia Universty, Fall 2022
  
  
• Guest Lecturer (on Dialogue Systems), Natural Language Processing
  Columbia Universty, Spring 2022
  
  
• Teaching Assistant, Conversational AI Special Topics (COMS 6998)
  Columbia University, Spring 2021

Services

• Area Chair, Women in Machine Learning 2022, NeurIPS 2022

• Publicity Chair, The 4th Workshop for Conversational AI, ACL 2022

• Conference reviewer: ACL, EACL, NAACL, EMNLP, *SEM, ICLR, AAAI, NeurlPS, ICML, CHI, NLPCC, ACL Rolling Review

• Journal reviewer: ACM Transactions on Human-Robot Interaction, Neurocomputing, ACM Transactions on Information Systems