Hi, my name is Huihan Li. I’m a final year PhD student working on Natural Language Processing in University of Southern California. I’m part of the INK Lab, advised by Xiang Ren. I got my M.S.E in Computer Science from Princeton University, being part of the Princeton Natural Language Processing Group and advised by Danqi Chen. I studied Computer Science and Cognitive & Linguistic Sciences at Wellesley College, working with Christine Bassem on human crowdsensing.

I am passionate about Natural Language Processing, Computational Linguistics, and everything about languages. In high school, I competed in International Linguistics Olympiad representing China, and won an Honorable Mention in Sofia, Bulgaria (2015) and a Bronze Medal in Mysore, India (2016).

Outside of research, I enjoy all kinds of sports, cooking/baking, and reading. I played water polo in college and it has been one of my best memories. Currently, I am learning tennis.

A functional, pain-free body is the foundation of all pursuits and endeavors. From my years of PT visits, I tried summarizing my learnings in a crash course on self-diagnosing and managing common bodily pains. I hope this is helpful for whomever is visiting my site ❤️.

Research

My research seeks to make large language models more transparent, reliable, and trustworthy reasoners. As language models increasingly rely on long reasoning traces to solve complex problems, their failures become difficult to diagnose. I study how to transform these opaque reasoning processes into auditable signals that reveal why models succeed or fail.

My work centers on three complementary questions:

• Knowledge. What information is a model recalling during reasoning, and how can we attribute its behavior to its pretraining experience?
• Coherence. Are intermediate reasoning steps logically supported by the available context and evidence?
• Utility. Which reasoning paths meaningfully contribute to solving a problem, and which merely increase reasoning length without improving outcomes?

Answering these questions requires both principled evaluation and new learning methods. My research develops techniques for diagnosing reasoning behavior, attributing model outputs to pretraining data, constructing challenging long-tail evaluation benchmarks, and exploring how diagnosed signals can be used to identify and steer reasoning errors.

Previously, I have worked on conversational question answering, constrained decoding, multilingual and multicultural evaluation, cultural bias in language models, and systematic generation of long-tail knowledge. A complete list of my publications can be found here.

News

• October 2025. I will be joining Adobe as an Applied Scientist Engineering Intern on the Agentic & GenAI team for the Adobe Experience Platform (AEP) starting in May 2026, where I will be working with Yunyao Li!
• August 2025. Our paper, “Diagnosing Memorization in Chain-of-Thought Reasoning, One Token at a Time”, is accepted to EMNLP 2025 Main Conference. See you in Suzhou in November!
• January 2025. I will be joining Meta GenAI as a Research Scientist intern starting May 2025, working with Bo Xiong!
• January 2025. Our paper, “Attributing Culture-Conditioned Generations to Pretraining Corpora”, is accepted to ICLR 2025. See you in Singapore in April!
• September 2024. Our paper, “In Search of the Long-Tail: Systematic Generation of Long-Tail Inferential Knowledge via Logical Rule Guided Search”, is accepted to EMNLP 2024 Main Conference. See you in Miami in November!
• August 2024. I am awarded the Amazon ML PhD Fellowship for 2024-2025. This fellowship will support my work on Secure and Trusted Machine Learning.
• July 2024. Our paper, “CULTURE-GEN: Revealing Global Cultural Perception in Language Models through Natural Language Prompting”, is accepted to COLM 2024. See you in Philly in October!
• March 2023. I will be joining AI2 Mosaic Team as a summer research intern starting May 2023, working with Nouha Dziri and Yejin Choi!

Experience

• Ink Lab, USC. PhD Student. Sept. 2022 - Present
  • Research in LLM reasoning, interpretability, and evaluation
  • Advisor: Xiang Ren
• Adobe, Adobe Experience Platform. Applied Scientist Engineering Intern. May 2026 - Aug. 2026
  • Research and engineering in agent routing and orchestration for enterprise agentic AI platforms
  • Mentors: Vineeth Mohan, Phu Tran, Kun Qian, Yunyao Li
• Meta, GenAI. Research Intern. May 2025 - Aug. 2025
  • Research in RL for Reasoning
  • Mentors: Bo Xiong, Liang Tan, Yipin Zhou, Derek Hao Hu
• AI2, Mosaic. Research Intern. May 2023 - March 2024
  • Research in Multicultural biases in LM
  • Mentors: Nouha Dziri, Yejin Choi
• Apple. AI/ML Intern. May 2022 - Aug. 2022
  • Individual NLP research/engineering project, Siri Information Intelligence, Answers and Web Ranking Team
  • Mentors & Supervisors: Michael Tu, Nihkil Ramesh, Chris Dubois
• Princeton NLP Group. M.S.E Student. Sept. 2020 - May 2022
  • Research in Natural Language Processing
  • Advisor: Danqi Chen
• Wellesley College. Research Assistant. Sept. 2018 - July 2020
  • Research in Mobile Crowdsensing
  • Advisor: Christine Bassem
• Google. SWE Intern. May 2019 - Aug. 2019
  • Individual engineering project, Shopping Assistant, Natural Language Team
  • Supervisors: Jesse Welch, John Karro

Teaching

• Machine Learning (CSCI 567). University of Southern California
• Foundations of Artificial Intelligence (CSCI 561). University of Southern California
• Introduction to Programming Systems (COS217). Princeton University
• Data Structures (CS230), Theory of Computation (CS235). Wellesley College

Honors and Awards

• Amazon Fellow. University of Southern California. Aug. 2024
• Siebel Scholars. Princeton University. Sept. 2021
• Sigma Xi Scientific Research Honor Society. Wellesley College. May 2020
• Durant Scholars magna cum laude. Wellesley College. May 2020

Service and leadership

• Reviewer. COLING 2024, ACL 2024, EMNLP 2024, ARR 2024, COLING 2025, ICLR 2025, ACL 2025, COLM 2025, EMNLP 2025, NeurIPS 2026.
• Student Representative on Board of Admission. Wellesley College. Oct. 2019 - May 2020