Liwei Jiang | 姜力炜 Ph.D. Candidate
Paul G. Allen School of Computer Science & Engineering, University of Washington
Contact: lwjiang@cs.washington.edu |
Google Scholar |
LinkedIn |
GitHub |
X
I am a Ph.D. candidate at Paul G. Allen School of Computer Science & Engineering, University of Washington, advised by Professor Yejin Choi. I am also a graduate research intern at NVIDIA and was previously a student researcher at Allen Institute for Artificial Intelligence (Ai2). I earned B.A. in Computer Science and B.A. in Mathematics from Colby College.
My research centers on humanistic AI safety, aiming to foster the synergistic, secure, and sustainable coexistence of AI and society—ultimately shaping the co-evolution of AI and humanity:
From Human to AI: Developing human-centered, continually-evolving, and future-oriented AI systems—drawing upon interdisciplinary insights into human intelligence, values, and global needs.
From AI to Human: Advancing the frontiers of knowledge about individuals and society, augmenting human capabilities, and addressing consequential sociotechnical challenges—through robust, efficient, and scalable innovations in data, learning algorithms, and system design.
My current research spans pluralistic alignment, self-improving algorithms for steerable and secure language models, anticipatory strategies for long-term risks—including overreliance and the erosion of human creativity—and socially beneficial applications.
Artificial Intelligence
AI Safety
Natural Language Processing
Machine Learning
Pluralistic Alignment
Human-AI Interaction
News
(Upcoming) Oct 2025: Co-organizing another edition of the SoLaR workshop on Socially Responsible Language Modelling Research at COLM 2025.
(Upcoming) July 2025: Our tutorial on Guardrails and Security for LLMs: Safe, Secure, and Controllable Steering of LLM Applications is accepted at ACL 2025. See you in Vienna!
March/May 2025: Give two guest lectures on Red-Teaming and Safeguarding Language Models: Current Practices, Challenges, and Future Directions at CMU and UCLA.
Oct/Nov 2024: Give two guest lectures on How to Build AI with Deep Concerns for Human Traits, Values, and Needs? at UIUC and University of Pittsburgh.
Oct/Nov 2024: Give two guest lectures on LLM Reasoning (In-Context Learning, Prompting, and Reasoning) at KAIST and UW.
Jan 2024: Lead of TA effort and co-design (a LLM version!) for the class module of CSE 447/517, the undergrad and grad NLP class at UW CSE.
Publications (*, + indicate equal contribution) Google Scholar
Preprints
Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)
Chasing Moving Targets with Online Self-Play Reinforcement Learning for Safer Language Models
X-Teaming: Multi-Turn Jailbreaks and Defenses with Adaptive Multi-Agents
AI Debate Aids Assessment of Controversial Claims
HAICosystem: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions
PolyGuard: A Multilingual Safety Moderation Tool for 17 Languages
WildHallucination: Evaluating Long-Form Factuality in LLMs with Real-World Entity Queries
2025
Investigating Machine Moral Judgment through the Delphi Experiment
Can Language Models Reason about Individualistic Human Values and Preferences?
CulturalBench: Building a Robust, Diverse, and Challenging Cultural Benchmark by Human-AI CulturalTeaming
Position Paper: Political Neutrality in AI is Impossible—But Here is How to Approximate it
SafetyAnalyst: Interpretable, Transparent, and Steerable Safety Moderation for AI Behavior
DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life
AI as Humanity's Salieri: Quantifying Linguistic Creativity of Language Models via Systematic Attribution of Machine Text against Web Text
To Err is AI: A Case Study Informing LLM Flaw Reporting Practices
2024
WildTeaming at Scale: From In-the-Wild Jailbreaks to (Adversarially) Safer Language Models
WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs
Value Kaleidoscope: Engaging AI with Pluralistic Human Values, Rights, and Duties
Culture-Gen: Revealing Global Cultural Perception in Language Models through Natural Language Prompting
Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement
Particip-AI: A Democratic Surveying Framework for Anticipating Future AI Use Cases, Harms and Benefits
Information-Theoretic Distillation for Reference-less Summarization
Position Paper: A Roadmap to Pluralistic Alignment
Impossible Distillation: from Low-Quality Model to High-Quality Dataset & Model for Summarization and Paraphrasing
JAMDEC: Unsupervised Authorship Obfuscation Using Constrained Decoding Over Small Language Models
MigraineTracker: Examining Patient Experiences with Goal-Directed Self-Tracking for a Chronic Health Condition
The Generative AI Paradox: "What It Can Create, It May Not Understand"
2023
What Makes it Ok to Set a Fire? Iterative Self-distillation of Contexts and Rationales for Disambiguating Defeasible Social and Moral Situations
NovaCOMET: Open Commonsense Foundation Models with Symbolic Knowledge Distillation
SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization
Reading Books is Great, But Not if You Are Driving! Visually Grounded Reasoning about Defeasible Commonsense Norms
Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuning
BiasX: "Thinking Slow" in Toxic Content Moderation with Explanations of Implied Social Biases
Faith and Fate: Limits of Transformers on Compositionality
ClarifyDelphi: Reinforced Clarification Questions with Defeasibility Rewards for Social and Moral Situations
2022
Aligning to Social Norms and Values in Interactive Narratives
Symbolic Knowledge Distillation: from General Language Models to Commonsense Models
NeuroLogic A*esque Decoding: Constrained Text Generation with Lookahead Heuristics
ProsocialDialog: A Prosocial Backbone for Conversational Agents
Quark: Controllable Text Generation with Reinforced Unlearning
2021
“I’m Not Mad”: Commonsense Implications of Negation and Contradiction
EnglishBot: An AI-Powered Conversational System for Second Language Learning
2019
QuizBot: A Dialogue-Based Adaptive Learning System for Factual Knowledge
Education & Experience
Education
University of Washington
Ph.D. in Computer Science and Engineering
Sept 2019 - Dec 2025 (expected), Seattle, Washington
Colby College
B.A. in Computer Science and Mathematics (Summa Cum Laude, Top 0.5%)
Sept 2015 - Dec 2018, Waterville, Maine
Professional Experience
NVIDIA
Research Intern at the Nemo-Guardrail Team
March 2025 - Present, Santa Clara, California
Allen Institute for Artificial Intelligence (Ai2)
Graduate Research Intern at the Mosaic and AllenNLP Teams
June 2020 - December 2024, Seattle, Washington
The Future Laboratory, Tsinghua University
Visit Student Researcher
July 2019 - September 2019, Beijing, China
Stanford University
Undergraduate Research Intern at the Computer Science Department
June 2017 - September 2019, Stanford, California
Honors & Awards
Outstanding Paper Award
CHI 2024 (May 2024)
Outstanding Paper Award
EMNLP 2023 (Dec 2023)
Best Paper Award
NAACL 2022 (Jul 2022)
Anne Dinning - Michael Wolf Endowed Regental Fellowship
University of Washington (2019)
Paul G. Allen School First-Year Ph.D. Fellowship
Member of the Phi Beta Kappa Society
Colby College (2018)
Elected as a member of Phi Beta Kappa with junior standing
Honorable Mention of Interdisciplinary Contest in Modeling (ICM)
COMAP (2018)
20th annual Interdisciplinary Contest in Modeling (ICM)
Phi Beta Kappa Undergraduate Scholastic Achievement Award
Colby College (2017)
Top two students in the sophomore and junior classes
Julius Seelye Bixler Scholar
Colby College (2016, 2017, 2018)
Top-ranking students as determined by the cumulative academic record, three-time recipient
Phi Beta Kappa Summer Research Scholar
Colby College (2016)
Summer research stipend