I am a faculty member of the School of Interactive Computing, Machine Learning Center, and NSF AI CARING Institute at Georgia Tech. My research lies at the intersections of machine learning, natural language processing, and social media. I direct the NLP X Lab which currently focuses on (1) analysis of large language models, such as cultural bias, multilingual capability, temporal shifts, and personalization; (2) text generation, such as constrained decoding and learnable evaluation metric; (3) NLP applications that can make impact in education, accessibility, etc. I recently received the NSF CAREER Award, Criteo Faculty Research Award, CrowdFlower AI for Everyone Award, Best Paper Award at COLING'18 and ACL'24, as well as research funds from DARPA and IARPA. I am a member of NAACL executive board. I was a postdoctoral researcher at the University of Pennsylvania. I received my PhD in Computer Science from New York University, MS and BS from Tsinghua University.
I'm recruiting 1-3 PhD students every year (apply to CS or Machine Learning PhD program and list me as a potential advisor; if you have EE background, consider also apply to ML ECE program). I recruit MS students (apply to MSCS program and email me) and undergraduates who have sufficient time and motivation for research theses.
Oct 2024, upcoming talk at Bloomberg's CTO Data Science Speaker Series
Sep 2024, talk at MIT on "Cultural Biases, World Languages, and Privacy Protection in Large Language Models"
Sep 2024, talk at Northeastern on "Human-AI Collaboration in Evaluating LLMs"
Mar 2024, talk at USC and UCLA on "Amazing Multilingual Capabilities and Concerning Cultural Biases in LLMs"
Dec 2023, invited talk on "Amplifying Multilingual LLMâs Cross-lingual Ability'' at the BrainLink event
Nov 2023, panel at WICT annual event on "How AI is used in Media, Entertainment and Technology"
Oct 2023, panel at HARC conference on "Simplifying Medical Texts with Large Language Models"
Oct 2023, demo of Thresh đŸ has been accepted to EMNLP 2023 -- a customizable tool for fine-grained human evaluation of LLM generated texts (e.g., MT, summarization, text revision, + more)
Aug 2023, I was quoted in Business Insider about AI-generated content online.
Aug 2023, Mounica Maddela defended her PhD thesis and will join Bloomberg AI's LLM group
Sep 2022, talk at Cornell Tech (video) "Importance of Data and Controllability in Neural Language Generation"
Aug 2022, my PhD student Mounica Maddela to start an internship at Meta AI; Yang Chen at Google Research.
Mar 2022, I received the NSF CAREER award! It will support my group's research on controllable text generation.
Research Highlights
Controllability, Stylistics, and Evaluation in Text Generation
We recently published one of the earliest works on formally evaluating the impressive text rewriting capability of GPT-3.5 and GPT-4, in particular, for paraphrase generation [EMNLPâ22a] and text simplification [EMNLPâ23a]. Our new LENS metric [ACLâ23a] is the first learned automatic evaluation metric for text simplification, which, when used as objective in minimum Bayes risk decoding (MBR), also set the newest state-of-the-art of open-sourced generation models, on par of GPT-3.5 and GPT-4. We also work on instruction-finetuning for style [ACLâ24a], edit-level text generation evaluation [EMNLPâ23a], document-grounded instructional dialog [ACLâ23b], document editing analysis for scientific writing [EMNLPâ22b].
Fairness, Multilingual, and Cross-cultural Capability of LLMs
We analyze monolingual and multilingual LLMs for cultural bias [ACLâ24b], distillation [ACLâ23c], cost efficiency [EMNLPâ21], robustness [ACLâ24d], and any other strengths/weaknesses that may lead to further development of better, fairer, smaller models. We also develop effective methods, such as label projection [ICLRâ24], for cross-lingual transfer learning. That is, with only English annotated data, we directly train multilingual language models that can perform tasks (e.g., entity recognition, question answer) in non-English languages.
NLP + X (social media, accessibility, privacy) Interdisciplinary Research
We work on a range of interesting and useful applications that aim to improve human life and society. A lot of our research has focused on text simplification [ACLâ23a,ACLâ23d,EMNLPâ21], which simplifies texts and improves readability, making knowledge accessible to all. We also recently started to develop document-grounded instructional dialog for personal assistance (e.g., cooking) [ACLâ23b], as part of the larger NSF AI CARING efforts. We also take a great interest in social media data, including work on human-in-the-loop detection of misinformation [ACLâ23e] and stance classification towards multilingual multi-cultural misinformation claims [EMNLPâ22b]. One of our current ongoing collaborative projects is looking at the privacy protection of users on social media.
I am a NAACL executive board member, a senior area chair for EMNLP 2024 (resource and evaluation), 2022 (generation), NAACL 2022 (machine learning for NLP), 2021 (generation), and ACL 2020 (generation), and an area chair for COLM 2024, ACL 2023 (semantics), EMNLP 2021 (computational social science), EMNLP 2020 (generation), AAAI 2020 (NLP), ACL 2019 (semantics), NAACL 2019 (generation), EMNLP 2018 (social media), COLING 2018 (semantics), EMNLP 2016 (generation), a workshop chair for ACL 2017, and the publicity chair for EMNLP 2019, NAACL 2018 and 2016. I also created a new undergraduate course on Social Media and Text Analytics.
Miscellaneous
When I have spare time, I enjoy visiting art museums, hiking, biking, and snowboarding.
photos together with Alan Ritter's group 
