Robin Jia

Profile picture

Email: robinjia at usc dot edu
Office: GCS 405E
Curriculum vitae

I am an assistant professor in the Thomas Lord Department of Computer Science at the University of Southern California, where I lead the AI, Language, Learning, Generalization, and Robustness (Allegro) Lab. My research seeks to understand modern deep learning systems for NLP and ensure that they are reliable. Some research questions I think about include:

  • How can we scientifically understand large language models? Our scientific understanding of LLMs lags far behind our ability to engineer them. To bridge this gap, we study mechanisms behind LLM capabilities such as in-context learning (NeurIPS 2024, EMNLP 2024), data memorization (NAACL 2024), and numerical reasoning (NeurIPS 2024).
  • How should we benchmark modern NLP systems? I have long advocated for benchmarking robustness (EMNLP 2017) and uncertainty (ACL 2020) of NLP systems. Our recent work has benchmarked generalization to long-tail examples (EACL Findings 2023) and calibration of LLMs (EMNLP Findings 2023). We have also shown that benchmarking under distribution shift can reveal advantages of neurosymbolic approaches (EMNLP Findings 2022).
  • How can we leverage LLMs when faced with complex reasoning tasks? We have developed methods that combine LLMs with symbolic solvers to solve complex long-term planning tasks (arXiv 2024). We have also improved smaller models on complex question answering tasks by training them to generate reasoning chains (EMNLP 2023).
  • How can advances in NLP inform other disciplines? On the legal side, we have collaborated with legal experts to operationalize underspecified requirements in the EU’s Digital Services Act (AIES 2024), and have developed techniques to prove that an LLM has been trained on a rightsholder’s data (Findings of ACL 2024). On the medical side, I collaborate with colleagues at the USC Keck School of Medicine to find ways to leverage LLMs for medical research and practice; previously, I built assisted curation tools for biomedical researchers (NAACL 2019).

I received my Ph.D. in Computer Science from Stanford University, where I was advised by Percy Liang. After that, I spent one year as a visiting researcher at Facebook AI Research, working with Luke Zettlemoyer and Douwe Kiela.

For prospective Ph.D. students

Unfortunately, I am not recruiting Ph.D. students for Fall 2025.

For USC undergraduate or master's students interested in research

If you are an undergraduate or master’s student at USC and are interested in doing research with me, please send me an email with the following:

  1. A description of why you’re interested in doing research.
  2. A summary of any experience you think may be relevant, including but not limited to coursework, previous projects, volunteer work, etc.
  3. A copy of your undergraduate and graduate (if applicable) transcripts.
  4. (Optional) Your CV.
For other undergraduate or master's students interested in research

Unfortunately, I do not have the bandwidth to advise undergraduate or master’s students from other universities at this time. One exception is the Viterbi Summer Undergraduate Research Experience (SURE) program. You are welcome to apply to this program and list me as a potential advisor.

News and Upcoming Events

Students

Ph.D. Students

Undergraduate and Masters Students

Lab Alumni

Publications

Why Do Some Inputs Break Low-Bit LLM Quantization?
Ting-Yun Chang, Muru Zhang, Jesse Thomason, and Robin Jia.
Workshop on Actionable Interpretability at ICML, 2025.

Verify with Caution: The Pitfalls of Relying on Imperfect Factuality Metrics.
Ameya Godbole and Robin Jia.
Findings of ACL, 2025.

Mechanistic Interpretability of Emotion Inference in Large Language Models.
Ala N. Tak*, Amin Banayeeanzade*, Anahita Bolourani, Mina Kian, Robin Jia, and Jonathan Gratch.
Findings of ACL, 2025.

Robust Data Watermarking in Language Models by Injecting Fictitious Knowledge.
Xinyue Cui, Johnny Tian-Zheng Wei, Swabha Swayamdipta, and Robin Jia.
Findings of ACL, 2025.

Interrogating LLM Design under a Fair Learning Doctrine.
Johnny Tian-Zheng Wei*, Maggie Wang*, Ameya Godbole, Jonathan H. Choi, and Robin Jia.
ACM Conference on Fairness, Accountability, and Transparency (FAccT), 2025.

Language Models can Infer Action Semantics for Classical Planners from Environment Feedback.
Wang Zhu, Ishika Singh, Robin Jia, and Jesse Thomason.
North American Association for Computational Linguistics (NAACL), 2025.
(github) (acl anthology) (bib)

TLDR: Token-Level Detective Reward Model for Large Vision Language Models.
Deqing Fu, Tong Xiao, Rui Wang, Wang Zhu, Pengchuan Zhang, Guan Pang, Robin Jia, and Lawrence Chen.
International Conference on Learning Representations (ICLR), 2025.

Pre-trained Large Language Models Use Fourier Features to Compute Addition.
Tianyi Zhou, Deqing Fu, Vatsal Sharan, and Robin Jia.
Neural Information Processing Systems (NeurIPS), 2024.
(github) (neurips page) (bib)

Transformers Learn Higher-Order Optimization Methods for In-Context Learning: A Study with Linear Models.
Deqing Fu, Tian-Qi Chen, Robin Jia, and Vatsal Sharan.
Neural Information Processing Systems (NeurIPS), 2024.
SoCalNLP Symposium 2023 Best Paper Award.
(github) (neurips page) (bib)

When Parts are Greater Than Sums: Individual LLM Components Can Outperform Full Models.
Ting-Yun Chang, Jesse Thomason, and Robin Jia.
Empirical Methods in Natural Language Processing (EMNLP), 2024.
(github) (blog) (acl anthology) (bib)

Operationalizing content moderation "accuracy" in the Digital Services Act.
Johnny Tian-Zheng Wei, Frederike Zufall, and Robin Jia.
AAAI/ACM Conference on AI, Ethics, and Society (AIES), 2024.

IsoBench: Benchmarking Multimodal Foundation Models on Isomorphic Representations.
Deqing Fu*, Ghazal Khalighinejad*, Ollie Liu*, Bhuwan Dhingra, Dani Yogatama, Robin Jia, and Willie Neiswanger.
Conference on Language Modeling (COLM), 2024.
(website) (dataset)

Proving membership in LLM pretraining data via data watermarks.
Johnny Tian-Zheng Wei*, Ryan Yixiang Wang*, and Robin Jia.
Findings of ACL, 2024.
(github) (acl anthology) (bib)

Do Localization Methods Actually Localize Memorized Data in LLMs?
Ting-Yun Chang, Jesse Thomason, and Robin Jia.
North American Association for Computational Linguistics (NAACL), 2024.
(github) (acl anthology) (bib)

Efficient End-to-End Visual Document Understanding with Rationale Distillation.
Wang Zhu, Alekh Agarwal, Mandar Joshi, Robin Jia, Jesse Thomason, and Kristina Toutanova.
North American Association for Computational Linguistics (NAACL), 2024.
(acl anthology) (bib)

Chain-of-Questions Training with Latent Answers for Robust Multistep Question Answering.
Wang Zhu, Jesse Thomason, and Robin Jia.
Empirical Methods in Natural Language Processing (EMNLP), 2023.
(github) (acl anthology) (bib)

SCENE: Self-Labeled Counterfactuals for Extrapolating to Negative Examples.
Deqing Fu, Ameya Godbole, and Robin Jia.
Empirical Methods in Natural Language Processing (EMNLP), 2023.
(github) (acl anthology) (bib)

Estimating Large Language Model Capabilities without Labeled Test Data.
Harvey Yiyun Fu, Qinyuan Ye, Albert Xu, Xiang Ren, and Robin Jia.
Findings of EMNLP, 2023.
(github) (acl anthology) (bib)

How Predictable Are Large Language Model Capabilities? A Case Study on BIG-bench.
Qinyuan Ye, Harvey Yiyun Fu, Xiang Ren, and Robin Jia.
Findings of EMNLP, 2023.
(github) (acl anthology) (bib)

Data Curation Alone Can Stabilize In-context Learning.
Ting-Yun Chang and Robin Jia.
Association for Computational Linguistics (ACL), 2023.
(github) (acl anthology) (bib)

Contrastive Novelty-Augmented Learning: Anticipating Outliers with Large Language Models.
Albert Xu, Xiang Ren, and Robin Jia.
Association for Computational Linguistics (ACL), 2023.
SoCalNLP Symposium 2022 Best Paper Award.
(github) (acl anthology) (bib)

Are Sample-Efficient NLP Models More Robust?
Nelson F. Liu, Ananya Kumar, Percy Liang, and Robin Jia.
Association for Computational Linguistics (ACL), 2023.
(acl anthology) (bib)

Do Question Answering Modeling Improvements Hold Across Benchmarks?
Nelson F. Liu, Tony Lee, Robin Jia, and Percy Liang.
Association for Computational Linguistics (ACL), 2023.
(acl anthology) (bib)

Does VLN Pretraining Work with Nonsensical or Irrelevant Instructions?
Wang Zhu, Ishika Singh, Yuan Huang, Robin Jia, and Jesse Thomason.
Open-Domain Reasoning Under Multi-Modal Settings Workshop at CVPR (O-DRUM), 2023.

Benchmarking Long-tail Generalization with Likelihood Splits.
Ameya Godbole and Robin Jia.
Findings of EACL, 2023.
(github) (acl anthology) (bib)

Generalization Differences between End-to-End and Neuro-Symbolic Vision-Language Reasoning Systems.
Wang Zhu, Jesse Thomason, and Robin Jia.
Findings of EMNLP, 2022.
(github) (acl anthology) (bib)

Knowledge base question answering by case-based reasoning over subgraphs.
Rajarshi Das, Ameya Godbole, Ankita Naik, Elliot Tower, Manzil Zaheer, Hannaneh Hajishirzi, Robin Jia, and Andrew McCallum.
International Conference on Machine Learning (ICML), 2022.
(github) (pmlr)

On the Robustness of Reading Comprehension Models to Entity Renaming.
Jun Yan, Yang Xiao, Sagnik Mukherjee, Bill Yuchen Lin, Robin Jia, and Xiang Ren.
North American Association for Computational Linguistics (NAACL), 2022.
(github) (acl anthology) (bib)

Models in the Loop: Aiding Crowdworkers with Generative Annotation Assistants.
Max Bartolo, Tristan Thrush, Sebastian Riedel, Pontus Stenetorp, Robin Jia, and Douwe Kiela.
North American Association for Computational Linguistics (NAACL), 2022.
(github) (acl anthology) (bib)

Question Answering Infused Pre-training of General-Purpose Contextualized Representations.
Robin Jia, Mike Lewis, and Luke Zettlemoyer.
Findings of ACL, 2022.
(github) (acl anthology) (bib)

Analyzing Dynamic Adversarial Training Data in the Limit.
Eric Wallace, Adina Williams, Robin Jia, and Douwe Kiela.
Findings of ACL, 2022.
(github) (acl anthology) (bib)

On Continual Model Refinement in Out-of-Distribution Data Streams.
Bill Yuchen Lin, Sida Wang, Xi Victoria Lin, Robin Jia, Lin Xiao, Xiang Ren, and Scott Yih.
Association for Computational Linguistics (ACL), 2022.
(website) (github) (acl anthology) (bib)

Dynaboard: An Evaluation-As-A-Service Platform for Holistic Next-Generation Benchmarking.
Zhiyi Ma*, Kawin Ethayarajh*, Tristan Thrush*, Somya Jain, Ledell Wu, Robin Jia, Christopher Potts, Adina Williams, and Douwe Kiela.
Neural Information Processing Systems (NeurIPS), 2021.
(website) (github) (blog post) (neurips page) (bib)

Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Little.
Koustuv Sinha, Robin Jia, Dieuwke Hupkes, Joelle Pineau, Adina Williams, and Douwe Kiela.
Empirical Methods in Natural Language Processing (EMNLP), 2021.
(github) (acl anthology) (bib)

Improving Question Answering Model Robustness with Synthetic Adversarial Data Generation.
Max Bartolo, Tristan Thrush, Robin Jia, Sebastian Riedel, Pontus Stenetorp, and Douwe Kiela.
Empirical Methods in Natural Language Processing (EMNLP), 2021.
(model) (github) (task) (acl anthology) (bib)

To What Extent do Human Explanations of Model Behavior Align with Actual Model Behavior?
Grusha Prasad, Yixin Nie, Mohit Bansal, Robin Jia, Douwe Kiela, and Adina Williams.
BlackBoxNLP Workshop, 2021.
(acl anthology) (bib)

The statistical advantage of automatic NLG metrics at the system level.
Johnny Tian-Zheng Wei and Robin Jia.
Association for Computational Linguistics (ACL), 2021.
(github) (acl anthology) (bib)

Evaluation Examples Are Not Equally Informative: How Should That Change NLP Leaderboards?
Pedro Rodriguez, Joe Barrow, Alexander Hoyle, John P. Lalor, Robin Jia, and Jordan Boyd-Graber.
Association for Computational Linguistics (ACL), 2021.
(website) (github) (acl anthology) (bib)

Do Explanations Help Users Detect Errors in Open-Domain QA? An Evaluation of Spoken vs. Visual Explanations.
Ana Valeria Gonzalez, Gagan Bansal, Angela Fan, Yashar Mehdad, Robin Jia, and Srinivasan Iyer.
Findings of ACL, 2021.
(acl anthology) (bib)

Swords: A Benchmark for Lexical Substitution with Improved Data Coverage and Quality.
Mina Lee*, Chris Donahue*, Robin Jia, Alexander Iyabor, and Percy Liang.
North American Association for Computational Linguistics (NAACL), 2021.
(github) (codalab) (acl anthology) (bib)

Dynabench: Rethinking Benchmarking in NLP.
Douwe Kiela, Max Bartolo, Yixin Nie, Divyansh Kaushik, Atticus Geiger, Zhengxuan Wu, Bertie Vidgen, Grusha Prasad, Amanpreet Singh, Pratik Ringshia, Zhiyi Ma, Tristan Thrush, Sebastian Riedel, Zeerak Waseem, Pontus Stenetorp, Robin Jia, Mohit Bansal, Christopher Potts, and Adina Williams.
North American Association for Computational Linguistics (NAACL), 2021.
(website) (github) (acl anthology) (bib)

N-ary relation prediction over text spans.
Hoifung Poon, Cliff Wong, and Robin Jia.
US Patent, 2021.

On the Importance of Adaptive Data Collection for Extremely Imbalanced Pairwise Tasks.
Stephen Mussmann*, Robin Jia*, and Percy Liang.
Findings of EMNLP, 2020.
(codalab) (github) (acl anthology) (bib)

With Little Power Comes Great Responsibility.
Dallas Card, Peter Henderson, Urvashi Khandelwal, Robin Jia, Kyle Mahowald, and Dan Jurafsky.
Empirical Methods in Natural Language Processing (EMNLP), 2020.
(github) (acl anthology) (bib)

Building Robust Natural Language Processing Systems.
Robin Jia.
Ph.D. Dissertation, 2020.

Selective Question Answering under Domain Shift.
Amita Kamath, Robin Jia, and Percy Liang.
Association for Computational Linguistics (ACL), 2020.
(codalab) (acl anthology) (bib)

Robust Encodings: A Framework for Combating Adversarial Typos.
Erik Jones, Robin Jia*, Aditi Raghunathan*, and Percy Liang.
Association for Computational Linguistics (ACL), 2020.
(codalab) (github) (acl anthology) (bib)

Certified Robustness to Adversarial Word Substitutions.
Robin Jia, Aditi Raghunathan, Kerem Göksel, Percy Liang.
Empirical Methods in Natural Language Processing (EMNLP), 2019.
(codalab) (github) (acl anthology) (bib)

MRQA 2019 Shared Task: Evaluating Generalization in Reading Comprehension.
Adam Fisch, Alon Talmor, Robin Jia, Minjoon Seo, Eunsol Choi, and Danqi Chen.
Workshop on Machine Reading for Question Answering (MRQA), 2019.
(github) (acl anthology) (bib)

Document-Level N-ary Relation Extraction with Multiscale Representation Learning.
Robin Jia, Cliff Wong, and Hoifung Poon.
North American Association for Computational Linguistics (NAACL), 2019.
(code and data) (acl anthology) (bib)

Know What You Don't Know: Unanswerable Questions for SQuAD.
Pranav Rajpurkar*, Robin Jia*, and Percy Liang.
Association for Computational Linguistics (ACL), 2018.
Best Short Paper Award.
(website) (codalab) (pptx slides) (pdf slides) (acl anthology) (bib)

Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfer.
Juncen Li, Robin Jia, He He, and Percy Liang.
North American Association for Computational Linguistics (NAACL), 2018.
(codalab) (pptx slides) (pdf slides) (acl anthology) (bib)

Adversarial Examples for Evaluating Reading Comprehension Systems.
Robin Jia and Percy Liang.
Empirical Methods in Natural Language Processing (EMNLP), 2017.
Outstanding Paper Award.
(codalab) (pptx slides) (pdf slides) (acl anthology) (bib)

Learning Concepts through Conversations in Spoken Dialogue Systems.
Robin Jia, Larry Heck, Dilek Hakkani-Tür, and Georgi Nikolov.
International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2017.
(data) (bib)

Data Recombination for Neural Semantic Parsing.
Robin Jia and Percy Liang.
Association for Computational Linguistics (ACL), 2016.
(codalab) (pptx slides) (pdf slides) (acl anthology) (bib)

"Reverse Genomics" Predicts Function of Human Conserved Noncoding Elements.
Amir Marcovitz, Robin Jia, and Gill Bejerano.
Molecular Biology and Evolution (MBE), 2016.

Mx1 and Mx2 Key Antiviral Proteins are Surprisingly Lost in Toothed Whales.
Benjamin A. Braun, Amir Marcovitz, J. Gray Camp, Robin Jia, and Gill Bejerano.
Proceedings of the National Academy of Sciences (PNAS), 2015.

* denotes equal contribution

Preprints

When Do LLMs Admit Their Mistakes? Understanding the Role of Model Belief in Retraction.
Yuqing Yang and Robin Jia.
arXiv, 2025.

Textual Steering Vectors Can Improve Visual Understanding in Multimodal Large Language Models.
Woody Haosheng Gan, Deqing Fu, Julian Asilis, Ollie Liu, Dani Yogatama, Vatsal Sharan, Robin Jia, and Willie Neiswanger.
arXiv, 2025.

Teaching Models to Understand (but not Generate) High-risk Data.
Ryan Wang, Matthew Finlayson, Luca Soldaini, Swabha Swayamdipta, and Robin Jia.
arXiv, 2025.

Cancer-Myth: Evaluating AI Chatbot on Patient Questions with False Presuppositions.
Wang Bill Zhu, Tianqi Chen, Ching Ying Lin, Jade Law, Mazen Jizzini, Jorge J. Nieva, Ruishan Liu, and Robin Jia.
arXiv, 2025.
(website) (dataset) (github)

Promote, Suppress, Iterate: How Language Models Answer One-to-Many Factual Queries.
Tianyi Lorena Yan and Robin Jia.
arXiv, 2025.

FoNE: Precise Single-Token Number Embeddings via Fourier Features.
Tianyi Zhou, Deqing Fu, Mahdi Soltanolkotabi, Robin Jia, and Vatsal Sharan.
arXiv, 2025.
(github) (website)

Rethinking Backdoor Detection Evaluation for Language Models.
Jun Yan, Wenjie Jacky Mo, Xiang Ren, Robin Jia.
arXiv, 2024.

Teaching

USC

Stanford

Professional Service

Other Work

Industry Internships

Undergraduate Research

Music

I have had the great pleasure of studying piano performance with Angela Wright and Laura Dahl. At various points, I have also studied solo piano with George Barth, duo piano with Kumaran Arul, and chamber music with Stephen Harrison.

Here are some of my recordings:

Piano duo concert, June 2017

Lisa Wang and I gave a piano duo concert on June 4, 2017.

* Probably not actually by Haydn

Piano Quintet Recital, May 2016

Ricky Wedeen, Brad Girardeau, Lee Fan, Andrew Guo, and I gave a concert on May 18, 2016, at which we performed the Schumann Piano Quintet, Op. 44 (mp3).

Senior Recital, April 2014

I gave my undergraduate senior recital on April 12, 2014. Here are the live audio recordings.

Other