ALLeGRo Lab

The AI, Language, Learning, Generalization, and Robustness (ALLeGRo) Lab is part of the Thomas Lord Department of Computer Science at the University of Southern California, led by Robin Jia. We study natural language processing and machine learning. Click on a research area to highlight our students working in that area:

News

Feb 2026 Johnny gave a talk at the Stanford NLP Seminar! Title: The Shape of AI Accountability and Its Contours in Copyright.
Jan 2026 Hubble and FoNE have been accepted to ICLR 2026!
May 2025 Deqing gave a talk at the Stanford NLP Seminar! Title: Closing the Modality Gap: Benchmarking and Improving Visual Understanding in Multimodal LLMs.

People

Faculty

Photo
Robin Jia
Assistant Professor
Website

PhD Students

Photo
Johnny Tian-Zheng Wei
PhD Student
Website
Hubble: a Model Suite to Advance the Study of LLM Memorization. ICLR 2026. [paper]
Photo
Ameya Godbole
PhD Student
Website
Analysis of Plan-based Retrieval for Grounded Text Generation. EMNLP 2024. [paper]
Photo
Wang Bill Zhu
PhD Student
Website
Efficient End-to-End Visual Document Understanding with Rationale Distillation. NAACL 2024. [paper]
Photo
Ting-Yun (Charlotte) Chang
PhD Student
Website
When Parts Are Greater Than Sums: Individual LLM Components Can Outperform Full Models. EMNLP 2024. [paper]
Photo
Deqing Fu
PhD Student
Website
Transformers Learn to Achieve Second-Order Convergence Rates for In-Context Linear Regression. NeurIPS 2024. [paper]
Photo
Tianyi Zhou
PhD Student
Website
Pretrained Large Language Models Use Fourier Features to Compute Addition. NeurIPS 2024. [paper]
Photo
Yuqing Yang
PhD Student
Website
When Do LLMs Admit Their Mistakes? Understanding the Role of Model Belief in Retraction. arXiv 2025. [paper]
Photo
Muru Zhang
PhD Student
Website
How Language Model Hallucinations Can Snowball. ICML 2024. [paper]

MS Students

Photo
Jerry Li
MS Student
Website LinkedIn
Are LLMs Reliable Rankers? Rank Manipulation via Two-Stage Token Optimization. arXiv. [paper]

Undergraduate Students

Photo
Eric Huang
Undergraduate Student

Alumni

Photo
Ryan Wang
Undergraduate → PhD at UC Berkeley
Website
Proving membership in LLM pretraining data via data watermarks. Findings of ACL 2024. [paper]
Photo
Lorena Yan
Undergraduate → PhD at Columbia University
Website
Promote, Suppress, Iterate: How Language Models Answer One-to-Many Factual Queries. EMNLP 2025. [paper]
Photo
Qilin Ye
Undergraduate → MS at Duke University
When Do Transformers Learn Heuristics for Graph Connectivity? [paper]
Photo
Harvey Yiyun Fu
Undergraduate → PhD at UChicago
Website
Estimating Large Language Model Capabilities without Labeled Test Data. Findings of EMNLP 2023. [paper]
Photo
Gustavo Lucas Carvalho
MS
LinkedIn

Projects

Hubble logo
Hubble
A suite of open-source LLMs for the scientific study of LLM memorization.
 allegro-project logo
allegro-project
A template for research projects at Allegro lab, with versioned experiments, shared library code, and SLURM integration.