Leon Li

profile_pic.jpg

al6843[at]nyu[dot]edu

CV Google Scholar LinkedIn GitHub HuggingFace Substack

Hi, I am Leon (Ang Li 李昂), a second year PhD student at New York University, advised by Prof. Pavel Izmailov and Prof. Micah Goldblum. Previously, I completed my B.S. and M.S. in Computer Science at Columbia University, where I was fortunate to work with Prof. Micah Goldblum, Prof. Tianyi Peng, Prof. Jing Dong, and Prof. Hongseok Namkoong.

My research focuses on three directions:

  • Learning and Memory: continual learning, memory consolidation, and end-to-end memory modules.
  • Long Context: context/KV-cache compression, test-time training, and long-horizon reasoning.
  • Science of LLMs: pretraining, synthetic data, and scaling laws.

News

Jun 26, 2026 New work on context compression! Covered by VentureBeat and Crypto Briefing. Models 🤗, code.
Sep 25, 2025 Two papers accepted to NeurIPS 2025!
Jul 22, 2025 Check out our new paper Zebra-CoT on interleaved text and visual reasoning! Dataset and model are available on Hugging Face 🤗.
May 01, 2025 I will be starting my PhD at NYU this Fall, working with Prof. Pavel Izmailov and Prof. Micah Goldblum!
Mar 24, 2025 Talk at AWS Responsible AI Group on LLM Agent Attacks.

Publications

2025

  1. lclm.png
    End-to-End Context Compression at Scale
    Ang Li*, Sean McLeish*, Haozhe Chen* , Nimit Kalra , Zaiqian Chen, Artem Gazizov, Venkata Anoop Suhas Kumar Morisetty, Bhavya Kailkhura, Harshitha Menon , Zhuang Liu, Brian R. Bartoldson, Tom Goldstein, Sanae Lotfi, Micah Goldblum, and Pavel Izmailov
    2025
    📄 Paper GitHub Code Hugging Face Model
  2. retrofitting.png
    Teaching Pretrained Language Models to Think Deeper with Retrofitted Recurrence
    Sean McLeish, Ang Li, John Kirchenbauer, Dayal Singh Kalra, Brian R. Bartoldson, Bhavya Kailkhura, Avi Schwarzschild, Jonas Geiping, Tom Goldstein, and Micah Goldblum
    2025
    📄 Paper GitHub Code Hugging Face Model
  3. zebra-cot.png
    Zebra-CoT: A Dataset for Interleaved Vision Language Reasoning
    Ang Li*, Charles Wang*, Deqing Fu*, Kaiyu Yue*, Zikui Cai*, Wang Bill Zhu*, Ollie Liu*, Peng Guo*, Willie Neiswanger, Furong Huang, Tom Goldstein, and Micah Goldblum
    2025
    📄 Paper GitHub Code Hugging Face Model Hugging Face Dataset
  4. digital-twin.png
    Twin-2K-500: A dataset for building digital twins of over 2,000 people based on their answers to over 500 questions
    Olivier Toubia, George Z Gui, Tianyi Peng, Daniel J Merlau, Ang Li, and Haozhe Chen
    Marketing Science, 2025
    📄 Paper GitHub Code Hugging Face Dataset
  5. Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attacks
    Ang Li*, Yin Zhou*, Vethavikashini Chithrra Raghuram*, Tom Goldstein, and Micah Goldblum
    2025
    📄 Paper
  6. persona.png
    LLM Generated Persona is a Promise with a Catch
    Ang Li, Haozhe Chen, Hongseok Namkoong, and Tianyi Peng
    Advances in Neural Information Processing Systems (NeurIPS), 2025
    📄 Paper Hugging Face Dataset
  7. permutation-invariant-overall.png
    Architectural and Inferential Inductive Biases For Exchangeable Sequence Modeling
    Daksh Mittal*, Ang Li*, Tzu-Ching Yen*, Daniel Guetta, and Hongseok Namkoong
    Advances in Neural Information Processing Systems (NeurIPS), 2025
    📄 Paper
  8. personallm_ds.png
    PersonalLLM: Tailoring LLMs to Individual Preferences
    Thomas Zollo*, Andrew Siah*, Naimeng Ye, Ang Li, and Hongseok Namkoong
    International Conference on Learning Representations (ICLR), 2025
    📄 Paper GitHub Code Hugging Face Dataset

2024

  1. qgym.png
    QGym: Scalable Simulation and Benchmarking of Queuing Network Controllers
    Haozhe Chen*, Ang Li*, Ethan Che*, Tianyi Peng, Jing Dong, and Hongseok Namkoong
    Advances in Neural Information Processing Systems (NeurIPS), 2024
    📄 Paper GitHub Code